in-situ studies of crystallization processes and other ...situation is the allotropism of carbon...
TRANSCRIPT
In-Situ Studies of Crystallization
Processes and Other Aspects of
Polymorphism
by
PHILIP ANDREW WILLIAMS
Thesis Submitted for
DOCTOR OF PHILOSOPHY
SCHOOL OF CHEMISTRY
CARDIFF UNIVERSITY
June 2014
i
DECLARATION This work has not been submitted in substance for any other degree or award at this or any other university or place of learning, nor is being submitted concurrently in candidature for any degree or other award. Signed ………………………………………… (candidate) Date ………………………… STATEMENT 1 This thesis is being submitted in partial fulfillment of the requirements for the degree of …………………………(insert MCh, MD, MPhil, PhD etc, as appropriate) Signed ………………………………………… (candidate) Date ………………………… STATEMENT 2 This thesis is the result of my own independent work/investigation, except where otherwise stated. Other sources are acknowledged by explicit references. The views expressed are my own. Signed ………………………………………… (candidate) Date ………………………… STATEMENT 3 I hereby give consent for my thesis, if accepted, to be available for photocopying and for inter-library loan, and for the title and summary to be made available to outside organisations. Signed ………………………………………… (candidate) Date …………………………
ii
Abstract
The work presented in this thesis represents the study of the polymorphism
exhibited by several molecular, organic solid-state systems. In-situ techniques are used
to explore aspects of the polymorphism and crystallization behaviour displayed by these
systems. The crystal structures of new polymorphs and other solid phases are
determined directly from powder X-ray diffraction data.
Chapter 1 provides background information on the phenomenon of polymorphism
and the importance of its study. In addition, the range of in-situ techniques that have
been used to study crystallization and solid-state systems is described.
Chapter 2 gives details on the experimental techniques used in the work presented
in this thesis. These include powder X-ray diffraction (including the methodology
behind structure determination from powder X-ray diffraction data), solid-state nuclear
magnetic resonance (NMR) spectroscopy, X-ray photoelectron spectroscopy, thermal
analysis techniques and dynamic vapour sorption.
Chapter 3 explores the polymorphism of m-aminobenzoic acid, resulting in the
discovery and characterization of three new polymorphs. The crystallization of
m-aminobenzoic acid from solution is studied via in-situ solid-state NMR spectroscopy
and this in-situ technique is expanded to allow the evolution of both the solid and liquid
phases during the crystallization process to be studied.
Chapter 4 investigates the crystallization behaviour of triphenylphosphine oxide
and methyldiphenylphosphine oxide from solution using in-situ solid-state NMR
spectroscopy. Evidence is found for transient, unknown solid phases formed during the
crystallization process.
Chapter 5 reports new insights into the crystallization behaviour of polymorph II
of the drug rac-ibuprofen. Differential scanning calorimetry and in-situ powder X-ray
diffraction experiments demonstrate that previous assumptions about the crystallization
behaviour of this polymorph were incorrect.
Finally, Chapter 6 explores the crystalline phases of the amino acid
L-phenylalanine. A new polymorph and two hydrate phases are discovered and their
hydration/dehydration behaviour characterized.
iii
Acknowledgements
Firstly, I would like to give my thanks to my supervisor, Professor Kenneth D. M.
Harris for all of his support, guidance and advice over the course of this work. I would
like to thank him for the opportunities he has given been me throughout this time and
the research he has allowed me to be a part of. I would also like to give very special
thanks to Dr. Colan Hughes for all of his excellent advice, help and support.
Furthermore, I truly give thanks to the other members of the research group both
past and present, Dr. Colan Hughes, Dr. Vasileios Charalampopoulos, Gregory
Edwards-Gau, Yuncheng Yan, Victoria Keast, Dr. Benjamin Palmer and Dr. Keat Lim
for always being ready to provide help or advice. It has been very enjoyable working
with all of you.
I would like to give my thanks to the UK 850 MHz Solid-State NMR Facility
(funded by EPSRC and BBSRC, as well as the University of Warwick including via
part funding through Birmingham Science City Advanced Materials Projects 1 and 2
supported by Advantage West Midlands and the European Regional Development
Fund) for use of their 850 MHz solid-state NMR spectrometer and Dr. Dinu Iuga for his
assistance while at the facility. I would also like to thank Professor Simon Gaisford and
Asma Buanz from University College London for their help with performing TGA and
DVS experiments.
Thanks to my friends (with special mention to Phil, Freya and Vic) and family for
their encouragement over the last few years.
Finally and most of all I would like to give my thanks to my parents. Without you
none of this would have been possible. Thank you both so much.
iv
Table of Contents
Chapter 1: Introduction
1.1 - Polymorphism ........................................................................................................... 1
1.1.1 - Importance of the Study of Polymorphism ........................................................ 5
1.2 - Crystallization ........................................................................................................... 7
1.3 - In-Situ Methods ........................................................................................................ 8
1.4 - Aims of this Project ................................................................................................ 12
Chapter 2: Experimental Methods
2.1 - X-Ray Diffraction ................................................................................................... 16
2.1.1 - Powder X-Ray Diffraction ............................................................................... 19
2.1.2 - X-Ray Sources ................................................................................................. 20
2.2 - Structure Determination from Powder X-Ray Diffraction Data ............................. 21
2.3 - Solid-State Nuclear Magnetic Resonance Spectroscopy ........................................ 26
2.4 - In-Situ Solid-State NMR Studies of Crystallization ............................................... 30
2.5 - X-Ray Photoelectron Spectroscopy ........................................................................ 31
2.6 - Differential Scanning Calorimetry.......................................................................... 33
2.7 - Thermogravimetric Analysis .................................................................................. 35
2.8 - Dynamic Vapour Sorption ...................................................................................... 36
Chapter 3: Discovery of a New System Exhibiting Abundant Polymorphism:
m-Aminobenzoic Acid
3.1 - Introduction ............................................................................................................. 37
3.2 - Experimental Methods ............................................................................................ 38
3.3 - Discovery and Characterization of New Polymorphs of m-ABA........................... 40
3.3.1 - Structural Description ...................................................................................... 44
3.3.2 - Stability ............................................................................................................ 53
3.4 - Details of Structure Determination ......................................................................... 56
3.4.1 - Structure Determination of Form III of m-ABA .............................................. 56
3.4.2 - Structure Determination of Form IV of m-ABA .............................................. 59
3.4.3 - Structure Determination of Form V of m-ABA ............................................... 61
3.5 - In-Situ Solid-State NMR Study of the Crystallization of m-ABA from Methanol 63
3.5.1 - Methodology of In-Situ Solid-State NMR Study ............................................. 63
3.5.2 - Results and Discussion ..................................................................................... 64
3.6 - In-Situ CLASSIC NMR Study of the Crystallization of m-ABA from DMSO ..... 66
3.6.1 - Methodology of CLASSIC NMR Experiment ................................................. 66
3.6.2 - Results .............................................................................................................. 67
3.7 - Conclusions ............................................................................................................. 74
v
Chapter 4: Discovery of New Solid Forms of Triphenylphosphine Oxide and
Methyldiphenyl-phosphine Oxide by In-Situ Solid-State NMR
4.1 - Introduction ............................................................................................................. 76
4.2 - Experimental Methods ............................................................................................ 78
4.2.1 - Crystallization within a Glass Capillary .......................................................... 80
4.3 - Results ..................................................................................................................... 81
4.3.1 - Triphenylphosphine Oxide ............................................................................... 81
3.4.2 - Methyldiphenylphosphine Oxide ..................................................................... 90
4.4 - Conclusions ............................................................................................................. 93
Chapter 5: New Insights into the Preparation of Form II of rac-Ibuprofen
5.1 - Introduction ............................................................................................................. 94
5.2 - Experimental Methods ............................................................................................ 95
5.3 - Results and Discussion ........................................................................................... 98
5.3.1 - DSC Studies of the Crystallization of Form II of rac-Ibuprofen ..................... 98
5.3.2 - In-Situ Powder XRD Studies of the Crystallization of Form II of rac-Ibuprofen
................................................................................................................................... 102
5.4 - Conclusions ........................................................................................................... 108
Chapter 6: Expanding the Solid-State Landscape of L-Phenylalanine: Discovery of
Polymorphism, New Hydrate Phases and Rationalization of
Hydration/Dehydration Processes
6.1 - Introduction ........................................................................................................... 109
6.2 - Experimental Details............................................................................................. 110
6.3 - Results and Discussion ......................................................................................... 112
6.4 - Details of Structure Determination from Powder XRD Data ............................... 123
6.4.1 - Structure Determination of L-Phe Hemihydrate ............................................ 123
6.4.2 - Structure Determination of L-Phe Monohydrate ............................................ 125
6.4.3 - Structure Determination of Polymorph II of L-Phe ........................................ 128
6.5 - Conclusions ........................................................................................................... 128
Chapter 7: Future Work ............................................................................................ 130
References .................................................................................................................... 133
Appendix A: Atomic Parameters for Crystal Structures Determined During This
Work ............................................................................................................................. 140
Appendix B: Full Differential Scanning Calorimetry Thermograms for
Experiments Performed on rac-Ibuprofen ............................................................... 149
Appendix C: Work Published during this PhD ........................................................ 157
1
Chapter 1: Introduction
1.1 - Polymorphism
The subject of polymorphism is both highly relevant and highly important to
many areas of science today. Polymorphism is the ability for a solid system to exist in
more than one crystalline form, with each form having identical chemical composition.
In general, these crystalline forms have differing physical properties, such as melting
point or solubility, and differing thermodynamic stability. Due to these differences in
stability, polymorphic transformations, in which one polymorph transforms into
another, may be observed. In some cases, however, kinetic factors may prevent the
transformation of a thermodynamically less stable form into a more stable form. In this
case, the less stable form is referred to as metastable. A famous example of this
situation is the allotropism of carbon (Figure 1.1). Under ambient conditions, -graphite
is more stable than diamond; however, kinetic factors prevent facile transformation,
allowing the metastable allotrope, diamond, to exist under ambient conditions.
It should be noted that while allotropism is often referred to as polymorphism of a
pure element, different allotropes may contain different molecular configurations of the
Figure 1.1: Crystal structures of (a) -graphite and (b) diamond.
2
element. In the case of -graphite and diamond, the carbon atoms obviously have
different covalent bonding schemes. This situation can also be seen in the case of
sulphur. While three allotropes, -orthorhombic, -monoclinic and -monoclinic
sulphur contain S8 molecules and have identical molecular composition, others contain
cyclo-Sn rings of other sizes or catena-Sn polymer chains. In contrast, in the case of
polymorphism, the molecular structure must be the same in each crystal structure (i.e.,
crystal structures containing isomers of a given molecule cannot be classed as
polymorphs). The term polymorphism is only applied when the multiple crystalline
forms have identical chemical composition. Thus, a hydrate or other solvate of a
compound is not a polymorph of that compound. Prototropic tautomerism (changes in
protonation state within the molecule) are, however, included within the definition of
polymorphism, allowing situations where, for example, the molecules in one polymorph
of a compound may be zwitterionic, while those in a second polymorph may be non-
zwitterionic.
Polymorphism is a fairly common phenomenon, though highly polymorphic
systems (containing more than three polymorphs) are rare. A famous example of a
highly polymorphic compound is 5-methyl-2-[(2-nitrophenyl)amino]-3-thiophene-
carbonitrile (Figure 1.2),1-3 known more commonly as ROY (named for the varying red,
orange and yellow colours of the different polymorphs). ROY has ten known
Figure 1.2: The polymorphs of 5-methyl-2-[(2-nitrophenyl)amino]-3-thiophene-
carbonitrile (ROY). Reprinted with permission from L. Yu, Acc. Chem. Res., 43, 1257-
1266 (2010). Copyright 2010 American Chemical Society.
3
polymorphs, though only seven of the polymorphs have known crystal structures.
Polymorphic forms may be related in one of two different ways: monotropically
or enantiotropically. In terms of thermodynamic definitions, an enantiotropic system is
one where the thermodynamic transition point (the point on a plot of free energy against
temperature at which the curves for each polymorph intersect, representing the
temperature at which the two polymorphs co-exist in equilibrium) is at a temperature
below the melting point of the lower melting polymorph. Similarly, a monotropic
system is one for which the thermodynamic transition point is at a temperature higher
than the melting point of the lower melting polymorph. These two types of relationship
are illustrated in Figure 1.3. If two forms are related monotropically, the transformation
of one to the other is irreversible (i.e., polymorph A can transform into polymorph B but
Figure 1.3: Plot showing how the Gibbs free energy of three polymorphs (termed I, II
and III) and the liquid phase of a compound may vary with temperature. Labelled are
the melting points of the three polymorphs (TI→L, TII→L, TIII→L), the temperature of the
transformation from I to II (TI→II) and the temperature of the transformation of I to III
(TI→III). I and II demonstrate an enantiotropic relationship as TI→II is below the melting
point of the two polymorphs. III demonstrates a monotropic relationship to the other
polymorphs as any transformation lies above the melting point of III.
4
polymorph B will never transform into polymorph A). If two forms are
enantiotropically related, the transformation between the two is reversible (i.e.,
polymorph A can transform into polymorph B and polymorph B can transform back
into polymorph A).
This situation can be demonstrated with the example of glycine (Figure 1.4).4-6 Of
Figure 1.4: The crystal structures of (a) -glycine, (b) -glycine and (c) -glycine.
Carbon atoms are shown in grey, hydrogen atoms in white, nitrogen atoms in blue and
oxygen atoms in red. Hydrogen bonds are shown as green dashed lines. (a) Also
includes a weak hydrogen bond formed between a hydrogen on the -carbon and an
oxygen from a neighbouring carboxylate group.
5
the three polymorphs of glycine found under ambient conditions (, and ), and
are enantiotropically related. The polymorph is the most stable at room temperature;
however, at high temperature this relationship is reversed, with a transition temperature
of ca. 165 °C for the transformation of →. The -polymorph, however, is
monotropically related to both and .
1.1.1 - Importance of the Study of Polymorphism
From the development of pharmaceuticals to the fundamental understanding of
the solid state, polymorphism is a vital area of study. First, we consider the role that
polymorphism plays in the pharmaceutical industry. As previously stated, different
polymorphs have differing physical properties. As a consequence, the solubility of any
particular drug will differ depending on the polymorph, which can have massive
consequence in terms of formulation and dosage. The most infamous example of the
problems caused by polymorphism in pharmaceuticals is the case of ritonavir [Figure
1.5(a)].7 Ritonavir, a protease inhibitor for the treatment of HIV, was developed by
Abbott Laboratories and formulated as both a liquid and a gel capsule. Ritonavir is not
bioavailable from the solid state and so both the liquid formulation and gel capsules
contained a solution in mixed ethanol/water. Approximately two years after
manufacture had begun, ritonavir was found to crystallize as a new phase within the
capsules. This phase was found to be a previously unknown polymorph with a much
greater stability than that of the previously known form. The solubility of this new
Figure 1.5: The molecular structures of (a) ritonavir, (b) rotigotine and (c) ranitidine
hydrochloride.
6
polymorph is much lower than that of the previously known form; the formulated
solution was not saturated with respect to the previously known form I, but was 400%
supersaturated with respect to the new form II. Traces of form II were able to cause
crystallization of the unusable form II and, thus, once the production facility became
contaminated with form II, manufacture of ritonavir had to be stopped. A reformulation
of ritonavir was necessary before production and sale could be restarted.
A second example8-10 of problems caused by the discovery of previously unknown
polymorphism of a pharmaceutical is the case of rotigotine [Figure 1.5(b)], a drug used
in the treatment of Parkinson’s disease and restless leg syndrome. Under the name
Neupro®, it was formulated as a transdermal patch containing rotigotine dispersed in an
adhesive coating. It was found, however, that rotigotine was crystallizing on the surface
of these patches, preventing absorption and therefore making them unsuitable for use.
These crystals were found to be a new polymorph, which resulted in Neupro® patches
being withdrawn from sale in the United States until reformulation was completed.
Polymorphism can also influence the patenting of pharmaceuticals. An example
concerns the drug ranitidine hydrochloride11,12 [Zantac, Figure 1.5(c)], a histamine H2-
receptor antagonist used to reduce stomach acid production. Ranitidine hydrochloride
was first patented by Glaxo in 1978 and a method for production was given. After scale
up to a plant process, one batch of the ranitidine hydrochloride produced was found to
have a different powder XRD pattern and IR spectrum from that described in the patent.
This material was found to be a new polymorph and named Form 2, with the original
form named Form 1. A new patent was granted for a process to produce Form 2
(claiming that Form 2 had preferable drying characteristics). Prior to expiry of the
patent for Form 1 (allowing marketing of generics), several companies, including
Novopharm, attempted to produce Form 1 but failed to do so, instead producing Form 2.
These companies attempted to invalidate the second patent by claiming that the method
used by Glaxo in the first patent inherently produced Form 2 and that the second patent
did not accurately describe the method used by Glaxo to produce Form 2 (omitting an
azeotroping process that allows for easier granulation of the product). Glaxo
successfully argued that the attempts by Novopharm had been contaminated by seed
crystals of Form 2 and went on to provide independent evidence of the production of
Form 1 by the method established in the first patent. Novopharm later went on to find a
method for reliable production of Form 1 and filed a new application to market this
7
product. Glaxo challenged this application, claiming that the product produced by
Novopharm would contain a mixture of Form 1 and Form 2 and so would infringe on
the second patent. Novopharm provided powder XRD patterns demonstrating no
detectable amounts of Form 2 in their product and so were allowed to market the
product.
Common pharmaceuticals that are known to be polymorphic include
paracetamol,13-15 aspirin16-20 and ibuprofen.21,22 Highly polymorphic pharmaceuticals
include the non-steroidal anti-inflammatory drug (NSAID) flufenamic acid,23 the
anticonvulsant carbamazepine24 and the antimicrobial sulfathiazole.25,26
1.2 - Crystallization
The crystallization of a solid from a supersaturated solution may be described in
terms of three major steps: pre-nucleation, the nucleation event and post-nucleation. At
the pre-nucleation stage, the solution is supersaturated with respect to the solubility of
the crystalline solid. The solution is therefore thermodynamically unstable and
crystallization will occur at some point (assuming that there is no insurmountable
kinetic barrier to the crystallization process). Initially, aggregation occurs between
solute molecules. The formation of an aggregate of a certain critical size (the smallest
size at which a particle is stable with respect to dissolution) corresponds to the
nucleation step. The size of stable crystal nuclei may range from as few as ten
molecules, to tens of thousands of molecules (nuclei of ice contain 80-100 molecules of
water).27 After the crystal nuclei have formed, crystal growth occurs. Crystal growth is
often faster along certain crystallographic axes, corresponding, for example, to the
direction of hydrogen bonded chains within a crystal structure. This unequal growth rate
results in the formation of a distinctive crystal morphology (e.g., hexagonal plates,
extended needles, etc.).
Crystal nucleation may be either primary or secondary. Primary nucleation occurs
in the absence of crystals, while secondary nucleation is induced by the presence of
other crystal nuclei. Primary nucleation may be subdivided into homogeneous and
heterogeneous types. Homogeneous nucleation occurs spontaneously within the solution
phase, while heterogeneous nucleation is induced by foreign particles or surfaces, or
more generally via the intervention of another phase in contact with the solution.
8
It has been suggested28 that polymorphism during crystallization from solution
may in part result from competition between multiple different aggregates. Each of
these aggregates would contain structural elements present in the final crystalline form.
A series of equilibria would be established between the free molecules, the various
molecular aggregates and the final crystals. Once nucleation has occurred, the solution
equilibria will be displaced to favour the fastest growing crystal nuclei. The fastest
nucleating and growing crystals may not be the most thermodynamically stable crystal
structure under these conditions and so can result in a metastable, kinetic product of
crystallization.
1.3 - In-Situ Methods
In-situ methods have been widely exploited in solid-state chemistry. Many
different experimental techniques have been used in these studies, including
spectroscopic methods, scattering methods and the definition could, perhaps, be
extended to thermal and gravimetric methods. The obvious advantage of an in-situ
method is the ability to observe the solid forms directly as a function of time during the
occurrence of some type of transformation, without removing it from the conditions
under which the transformation occurs. This can allow insight into the presence, for
example, of transient intermediate phases during crystallization processes or forms
which are simply not stable under ambient conditions and so could never be removed
for ex-situ analysis.
In-situ studies using scattering methods have shown some success, especially in
the field of inorganic chemistry. Much work has been undertaken using both
conventional (or angular dispersive) X-ray diffraction (XRD) and energy dispersive
XRD.29 In general, however, greater success has been found using energy XRD as this
uses a polychromatic X-ray beam (from a synchrotron source) which is of higher
intensity than the monochromatic beams used in conventional XRD, allowing use of
larger and more elaborate in-situ vessels (as absorption of the X-ray beam is less
problematic due to the higher intensity). The study of the formation of the microporous
gallophosphate ULM-530 is an excellent example where both conventional and energy
dispersive XRD have been combined to fully understand the processes occurring during
crystallization. Through use of both techniques, two new intermediates were found
during the crystallization of ULM-5 and the conditions that favour one over the other
9
were discovered. A second example is the use of powder XRD to study the solid-state
reaction of CaCO3 and TiO2 to form CaTiO3, a dielectric material.31 By taking repeated
powder XRD patterns while heating a mixture of the starting materials, it was found that
mechanical activation greatly enhanced the rate of reaction and allowed the production
of a phase-pure product.
In-situ studies using neutron diffraction methods have also been used extensively.
There are, of course, both advantages and disadvantages of using neutron diffraction in
comparison to XRD for in-situ studies. Among the advantages is the ability to select a
material for the reaction vessel with a low neutron absorption cross-section, rendering
the vessel essentially invisible during the diffraction measurements. As a consequence,
larger reaction vessels can be used, including autoclaves, cryostats or pressure vessels.
In addition, due to the fact that neutrons are scattered by the atomic nucleus, rather than
the electrons as in the case of XRD (the theory behind XRD is described in Chapter 2),
light elements that are normally difficult to observe via XRD are easily observed via
neutron diffraction. This aspect may be highly advantageous in, for example, studies
involving water (such as the crystallization of hydrates). Disadvantages of the use of
neutron diffraction include the low flux of neutrons produced by neutron sources
(compared to X-ray sources) combined with the inability to collimate neutron beams,
necessitating large sample volumes and long acquisition times. As a consequence, both
short-lived intermediates and intermediates only produced in small amounts during the
in-situ study may not be observable by neutron scattering methods.
For diffraction to occur, the scattering from a material must be coherent and
elastic, which is an important factor in neutron diffraction studies of samples containing
hydrogen. The high natural abundance (99.98%)32 1H isotope of hydrogen scatters
mostly incoherently, however, the 2H (deuterium) isotope, scatters mostly coherently.
Thus, in order to acquire high quality neutron diffraction data, deuterated samples must
be used. Unfortunately, deuteration may be limited by difficulties in synthesis (e.g., it
may be impossible to produce deuterated versions of a naturally occurring mineral
sample) or it may be prohibitively expensive. An early example33 of in-situ neutron
diffraction is the study of the intercalation of ammonia into tantalum disulphide
(specifically, the polymorph 2H-TaS2). Through use of neutron diffraction, it was found
that the intercalation occurs with two different reaction periods, first a slow nucleation
process, followed by faster growth.
10
For spectroscopic methods, many experimental techniques have been used for in-
situ studies, ranging from Fourier transform infra-red (FTIR) and Raman spectroscopy,
to X-ray absorption spectroscopy (XAS) and nuclear magnetic resonance (NMR)
spectroscopy. Raman spectroscopy has been used, in situ, to study the solvent-mediated
polymorphic transformation of L-glutamic acid.34 During this experiment, Raman
spectroscopy was used to monitor the polymorphic form present, while attenuated full
reflection Fourier transform infrared (ATR-FTIR) spectroscopy was used to study the
concentration in the solution phase. In addition, focused beam reflectance measurement
(FBRM) and an optical imaging technique were used to track the chord length and
morphology of the crystals, respectively. This combination of techniques allowed the
characterization of the nucleation and growth of the two polymorphic forms (the
metastable form and stable form), and also allowed the study of the dissolution of
the metastable polymorph. Raman spectroscopy has been coupled with dynamic vapour
sorption (DVS) measurements for a tandem in-situ study of the hydration of an active
pharmaceutical ingredient.35 Raman spectroscopy has also been used to great effect for
the in-line monitoring of pharmaceutical production processes.36
Solid-state NMR spectroscopy has many advantages in relation to in-situ studies
of crystallization but the most important advantage compared to other spectroscopic
techniques is the ability to selectively observe only the solid state, completely excluding
the solution state from the measurement. Certain challenges must, however, be met
when attempting in-situ solid-state NMR studies of crystallization, the most practical of
which is finding a method of sealing a liquid sample within a solid-state NMR rotor
spinning under magic-angle conditions (typically with a spinning frequency of 8 to 12
kHz). Details on how this challenge may be overcome are given in Chapter 2. In
addition, the system being studied must involve suitable nuclei for solid-state NMR
spectroscopy. Examples include 13C and 31P (though due to the low natural abundance
of 13C, a labelled sample may be required). This technique has previously been used
successfully to study crystallization of the amino acid glycine from solution. Glycine
has five known polymorphs37-41 termed , , , and (though only , and have
been found under ambient conditions, with and formed only under high pressure).
Under ambient conditions, is the most stable form, followed by and then .
Crystallization of glycine from H2O (with natural isotopic abundance) generally gives
the -form. However, when crystallized from D2O, formation of the -form is favoured.
This apparent isotope effect on the resultant polymorph of the crystallization was
11
studied by in-situ solid-state NMR.42 It was found that during crystallization from H2O,
-glycine is quickly formed followed by no further change in the solid-state (within ca.
14 hours). When crystallizing from D2O, -glycine is again formed initially, but then
transforms over the course of several hours into the form. This observation
demonstrates that the difference when crystallizing from the H2O and D2O is not in the
initial nucleation process (the polymorph is produced initially in crystallization from
both H2O and D2O) but instead the fact that D2O catalyses the polymorphic
transformation of -glycine to -glycine. The in-situ NMR technique was also applied
to study crystallization of -glycine from a water/methanol mixture.43 It was found that
after adding methanol to an aqueous solution of glycine, the resultant anti-solvent
crystallization initially produces -glycine, which then quickly transforms into the more
stable -polymorph. In-situ solid-state NMR has also been applied to inorganic systems,
such as crystallization of zeolite A from a gel.44 This in-situ measurement involved both
27Al and 29Si direct-excitation measurements (under magic-angle spinning of 2.1 kHz)
to chart the progress of the crystallization.
Thermal analysis techniques may be considered as a type of in-situ study,
including such techniques as thermogravimetric analysis (TGA) and differential
scanning calorimetry (DSC). DSC has been combined with small and wide angle X-ray
scattering (SAXS and WAXS, respectively) in the study of the polymorphic transition
of the lipid trilaurin45 (the triglyceride of lauric acid). Trilaurin with a 4% addition of
cholesterol was measured by DSC after tempering at varying temperature. During the
DSC measurement, SAXS and WAXS patterns were acquired simultaneously. A
synchrotron radiation source was used to minimize the amount of time required to
acquire the X-ray scattering data (a heating rate of 1 °C/min was used to acquire the
data with each set of X-ray scattering data acquired over approximately one minute).
As previously stated, each in-situ technique has particular advantages and
disadvantages that must be assessed when designing an experimental protocol. When
the system to be studied involves fast changes and quick evolution, a method that allows
for maximal time resolution is essential. In these cases, a spectroscopic method such as
IR, Raman or NMR may be preferred. To allow for optimal time resolution during an
in-situ NMR study, two of the three following conditions must generally be met: the
sample must have fast relaxation behaviour, the nucleus to be studied must either have
high natural abundance or be artificially enriched, or a high field instrument must be
12
used. When the studied process is slower but maximal structural information is desired,
in-situ powder XRD or powder neutron diffraction may be considered as the best
choice. Through use of either of these methods, the crystal structures of intermediate
solid phases may be determined. It should also be noted that in-situ NMR spectroscopy
studies the entire sample, while other spectroscopic and diffraction techniques generally
only interrogate a small part of the sample (e.g., the location of the incident beam in
XRD).
1.4 - Aims of this Project
One of the aims of this project has been to develop in-situ techniques and to use
them in combination with ex-situ techniques to gain the maximum insight possible from
solid-state systems, including the further development of the in-situ NMR technique
previously used in the study of the crystallization of glycine, as mentioned above. These
in-situ studies have been combined with ex-situ analysis and structure determination
from powder XRD data. Other in-situ techniques used in this project include in-situ
powder XRD. The systems studied involve simple organic molecules, with similarity to
active pharmaceutical compounds (and, in the case of ibuprofen, a very widely used
drug).
The aminobenzoic acids are a class of simple organic compound used as
feedstocks in the production of many important molecules. All three isomers,
o-aminobenzoic acid, m-aminobenzoic acid and p-aminobenzoic acid exhibit
polymorphism. As the aminobenzoic acids contain both an acidic carboxylic acid group
and a basic amine group, the possibility exists for the molecules to form zwitterions.
o-Aminobenzoic acid (o-ABA) has three known polymorphs designated I,46 II47 and
III.48 Forms II and III both contain non-zwitterionic molecules of o-ABA, while Form I,
interestingly, contains both zwitterionic and non-zwitterionic molecules of o-ABA.
Prior to this work, m-aminobenzoic acid (m-ABA, Figure 1.6) was known to have two
polymorphs, Forms I and II.49 Form I of m-ABA was known to contain only
zwitterionic molecules of m-ABA,50 while Form II contains only non-zwitterionic
molecules of m-ABA. Only the crystal structure of Form II51 was known prior to the
current work, presumably due to the inability to grow single crystals of sufficient size
and quality to solve the crystal structure of Form I using single-crystal X-ray
diffraction. p-Aminobenzoic acid is known to have two polymorphs, and .53 Both
13
polymorphs contain only non-zwitterionic molecules of p-ABA. It was previously
believed that there were four polymorphs of p-ABA.54 However, it was discovered that
two of these materials are actually solvates with acetone and 1,4-dioxane.55 Within this
project, the polymorphism of m-aminobenzoic acid has been studied, revealing the
abundant polymorphism of this system.
The second system studied in this project is triphenylphosphine oxide [TPPO,
Figure 1.7(a)]. TPPO is commonly encountered in the field of organic chemistry. It is
produced as a by-product from several synthesis reactions including the Mitsunobu
reaction, Staudinger reaction and Wittig reaction. TPPO is also used as a ligand and as a
crystallization aid.56 Both TPPO and its cocrystals tend to form large blocky crystals.
TPPO is known to have four polymorphs: monoclinic I,57 orthorhombic,58 monoclinic
II59 and monoclinic III. While crystal structures have been reported for all four
polymorphs, only two (monoclinic I and orthorhombic) have known and repeatable
methods of production. Monoclinic II was found serendipitously during the attempted
crystallization of a metal complex and the structural information on monoclinic III in
the CSD database was provided only as a “private communication” with no information
Figure 1.6: The molecular structure of m-aminobenzoic acid as (a) the non-zwitterionic
tautomer and (b) the zwitterionic tautomer.
Figure 1.7: The molecular structures of (a) triphenylphosphine oxide and (b)
methyldiphenylphosphine oxide.
14
on the method of production. During this work, the crystallization of TPPO has been
studied via in-situ solid-state 31P NMR, revealing the presence of new transient solid
phases.
The related compound, methyldiphenylphosphine oxide [MDPPO, Figure 1.7(b)],
was not previously known to be polymorphic and had a single known crystal structure.60
As in the case of TPPO, MDPPO finds use as a ligand and also in a variety of organic
syntheses.61-63 The crystallization of MDPPO is also studied in this work by in-situ 31P
NMR and evidence for new transient phases found. This suggests that MDPPO may
indeed be polymorphic.
Ibuprofen (Figure 1.8), as previously mentioned, is a widely used non-steroidal
anti-inflammatory drug. Ibuprofen is most commonly used as the racemic free acid
(rac-ibuprofen) which has recently been discovered to be polymorphic22 and the crystal
structure of the new polymorph has been determined from powder XRD data.64
Ibuprofen is most commonly found as Form I, the most stable form at room
temperature. Molten rac-ibuprofen (Form I melting point: 349 K) can easily be
supercooled and is found to undergo a glass transition at approximately 228 K. Form II
is reported22 to be produced when molten rac-ibuprofen is quench cooled to a
temperature at least 60 K below the glass transition point, before annealing for several
hours at 258 K. Form II of rac-ibuprofen has a melting point of 290 K, a value 59 K
below that of Form I. During this work, DSC and in-situ powder XRD were used to
explore the conditions required to produce Form II of rac-ibuprofen, to discover
whether the quench to a temperature below the glass transition is essential.
L-Phenylalanine (L-Phe, Figure 1.9) is one of the 20 directly genetically encoded
amino acids. Industrially, L-Phe is used in the production of the artificial sweetener
aspartame [N-(L-α-aspartyl)-L-phenylalanine, 1-methyl ester]. L-Phe is known,
industrially, in both an anhydrous solid form65 and a hydrous solid form.66 In this
context, crystallization as the anhydrate is preferred, as the hydrate exhibits a long
Figure 1.8: The molecular structure of ibuprofen.
15
needle-like morphology that makes processing very difficult.66 Within the literature,
there has been confusion over the existence of polymorphism within the L-Phe system.
There are multiple entries within the Cambridge Structural Database (CSD) for
anhydrous L-Phe. However, only one of these entries contains atomic coordinates (this
structure was actually determined using D-Phe), corresponding to the commonly found
anhydrous form of L-Phe.65 Work67 studying the crystallization product of L-Phe from
water and a mixed water/ethanol solvent by solid-state 13C NMR found that different
crystalline forms were obtained from each solvent. During this project, crystallization
experiments have been carried out to produce new hydrate phases and the hydration
behaviour has been analysed using dynamic vapour sorption (DVS) measurements
which, combined with in-situ powder XRD, have led to the discovery of a new
polymorph of anhydrous L-Phe.
Figure 1.9: The molecular structure of L-phenylalanine.
16
Chapter 2: Experimental Methods
2.1 - X-Ray Diffraction
X-ray diffraction (XRD) is a very powerful and widely used technique for the
analysis of materials. It is generally non-destructive (though sensitive samples can
suffer radiation damage and some sample containment techniques may result in an
unrecoverable sample) and can provide a large amount of information about the sample.
The vast majority of crystal structures are solved using single-crystal XRD, where
three-dimensional diffraction data for a crystal is recorded. Powder XRD, on the other
hand, tends to be used for more analytical purposes: determining the identity of a
crystalline sample, analysis of crystalline domain size and in-situ study of
crystallizations and transformations. However, advances over the past 20 years now
allow crystal structures to be determined directly from powder XRD data and the
procedures used in such work are discussed more in Section 2.2.
X-ray waves are scattered by the electrons within atoms. Each atom within a
crystal will scatter X-ray waves in all directions. A regularly spaced array of scatterers
(such as the atoms within a crystal) will result in a regularly spaced array of diffracted
waves. In most directions, destructive interference causes the X-ray waves to be
“cancelled out”, however, in specific directions, constructive interference will occur and
a diffraction spot will be produced. For constructive interference to occur (and, thus for
a diffraction maximum to be observed), the diffracted waves must be in phase with one
another and so the path difference between X-ray waves must be an integer multiple of
the wavelength.
Bragg demonstrated68 that the angular distribution of diffraction maxima in a
diffraction pattern can be rationalized by treating the scattered X-rays as if they are
reflected from planes passing through points of the crystal lattice (Figure 2.1). This
reflection is analogous to that of light from a mirror, thus, the incident angle of the
X-ray is equal to the angle of reflection. As mentioned above, for two waves “reflected”
from parallel planes to be in phase, the path difference between the two (AB + BC)
must be an integer multiple of the wavelength (Equation 2.1).
nBCAB . (2.1)
17
As seen in Figure 2.1:
sindBCAB , (2.2)
which, combined with Equation 2.1 results in Bragg’s Law:
sin2 hkldn , (2.3)
where n is the order of the diffraction maximum, is the wavelength of incident
radiation, dhkl is the spacing between a particular set of lattice planes defined by the
Miller indices hkl and is the angle between the incident X-ray beam and the lattice
plane (corresponding to half the diffraction angle 2).
The intensity, Ihkl, of the Bragg reflection with Miller indices hkl, is proportional
to the square of the magnitude of the structure factor, Fhkl:
2
hklhkl FI . (2.4)
The structure factor, Fhkl, is defined by the following equation:
n
j
jlzkyhxi
jjhkl
BeNfF jjj
12
2
2 sinexp
, (2.5)
where the summation is over all atoms in the unit cell, fj is the atomic scattering factor
for the jth atom, Nj is the site occupancy for the jth atom, xj, yj and zj are the atomic
Figure 2.1: Derivation of Bragg’s Law from the scattering of X-rays from parallel
lattice planes within a crystal structure.
18
coordinates for the jth atom and Bj is the isotropic displacement factor (Biso) for the jth
atom, a measure of atomic displacement (which may be caused by a variety of effects
both temperature dependent and temperature independent, including thermal vibration
of atoms, dynamic disorder and static disorder). Biso may be related to the also
commonly used Uiso:
28
isoiso
BU (2.6)
The use of Uiso allows the determination of physically relevant properties as
2uU iso , (2.7)
where u is the atomic displacement and represents averaging over space and time.
Thus, the square root of Uiso gives a root mean square value for the atomic
displacement.
Each atomic scattering factor can be calculated through the equation:
4
1
2
2sinexp
i
ii
bacf
, (2.8)
where ai, bi and c are a set of coefficients determined from a fit to a simulated atomic
scattering curve69 (calculated, depending on the atom or ion concerned, by a number of
methods including relativistic Hartree-Fock and Dirac-Slater wavefunctions) and in this
way depend on the electronic structure of the atom. Scattering of X-rays by an atom
increases with atomic number (i.e., as the number of electrons within the atom
increases). Consequently, elements with higher atomic number scatter X-rays more
strongly than elements of lower atomic number. Hydrogen atoms, with only a single
electron, can therefore be difficult to observe amongst the scattering produced by
heavier atoms (even against carbon, oxygen and nitrogen, as commonly found in simple
organic molecules). As described later, locating hydrogen atoms during structure
determination from X-ray diffraction data can be difficult for this reason. Other methods
may be required for accurate determination of hydrogen atom positions and for this
purpose neutron diffraction is often performed. The scattering of neutrons from an atom
does not depend upon the electrons but instead on the nuclear structure and so there is
no overall increase in scattering power with increasing atomic number as seen for X-ray
diffraction.
19
2.1.1 - Powder X-Ray Diffraction
A powder sample is ideally made up of a large number of randomly oriented
crystallites. In the ideal situation, there is a random distribution of crystallite
orientations within the powder. Due to this random distribution, unlike the diffraction
“spots” produced in the case of single-crystal XRD, powder XRD produces diffraction
cones. As a consequence of the fact that diffraction data are effectively compressed into
one dimension in powder XRD (compared to the three dimensions in single-crystal
XRD), there is usually significant overlap of diffraction peaks in the powder XRD
pattern. This is especially true for materials with large unit cells and/or low symmetry
(as found for most molecular solids), where there is a very high density of peaks in the
powder XRD pattern. The process for structure solution from powder XRD data is
further explained in Section 2.2.
As may be expected, actual powder samples may not conform to the ideal
situation mentioned above and will not have an entirely random distribution of
crystallite orientations. Crystallites will have a particular morphology, including shapes
such as plates or needles. Due to this, crystallites will often lay preferentially in certain
orientations, for example, plates stacking flat on top of one another other or needles
aligning along the same direction (Figure 2.2). The non-random distribution of
orientations causes the measured relative peak intensities to differ from their intrinsic
values and so can cause difficulty during structure determination. This preferred
orientation effect, when severe, can prevent structure solution from powder XRD data
and so, therefore, effort must be made to reduce or remove it. Preferred orientation is
less pronounced in samples contained within glass capillaries than those held flat
between pieces of tape within a foil type sample holder. Preferred orientation may be
reduced further by mixing the sample with an amorphous phase (amorphous to avoid
observing diffraction peaks from the second phase).
Two geometries were used during the course of this project, Bragg-Brentano
reflection and -2 transmission. Transmission geometry was utilized with a foil-type
sample holder, with samples either held flat between two pieces of tape or contained
within sealed glass capillaries. Transmission geometry has the advantages of superior
line shapes, reduced preferred orientation and higher instrumental resolution over
reflection geometry making the data more suitable for structure determination.
20
Reflection geometry was used for variable temperature powder XRD measurements
under vacuum using a Phenix cryostat.
2.1.2 - X-Ray Sources
X-rays for use in diffraction experiments may be generated in different ways.
Laboratory sources are based on an X-ray tube, in which electrons are emitted from a
heated filament cathode and are accelerated through a high voltage towards a metal
target anode. The electrons striking the anode cause excitation and emission of core
electrons from the K shell (1s orbital). Electrons from either the L or M shell
(specifically, the 2p or 3p orbitals respectively) relax to the K shell, producing either K
or K radiation, respectively. In laboratory experiments, Cu K1 radiation is typically
used for powder diffraction (the K radiation is actually made up of two components,
K1 and K2 due to the difference in energy between the spin states of the 2p orbital).
The K1 radiation is specifically selected through use of a single-crystal
monochromator made of a material such as germanium.
X-rays may also be generated by a synchrotron. In a synchrotron, a beam of
electrons is accelerated through use of a linear accelerator before injection into a large
storage ring, around which the electrons travel at constant energy. At specific points, the
beam of electrons is bent, causing an acceleration resulting in the emission of X-rays.
These X-rays are of very high intensity, are highly collimated, are tuneable (i.e., all X-
ray wavelengths are available and any specific wavelength may be selected through use
of an appropriate monochromator), are highly polarized and the radiation is pulsed.
Figure 2.2: Schematic demonstrating (a) a powder consisting of randomly oriented
crystallites and (b) a powder in which the crystallites exhibit substantial preferred
orientation.
21
Instrumental line broadening effects are reduced when using synchrotron radiation,
resulting in high-intensity and high-resolution data.
2.2 - Structure Determination from Powder X-Ray Diffraction Data
As mentioned previously, the powder XRD pattern is related to the crystal
structure through the structure factor (Equations 2.2 and 2.4), allowing the powder XRD
pattern corresponding to a known crystal structure to be calculated easily. However, the
reverse process, the determination of a crystal structure from a diffraction pattern is
much more complicated. The crystal structure may be represented in the case of XRD
by the electron density (r) within the unit cell. Atoms will obviously be represented by
maxima within an electron density map of the unit cell.
By defining the structure factor for a diffraction maximum as F(H), where H is
the scattering vector in reciprocal space, *c*b*aH lkh , and a*, b* and c* are
the reciprocal lattice vectors, it is related to the electron density as follows:
rrHrHHH diiFF )2exp()()](exp[)()( , (2.9)
where the structure factor F(H) has magnitude )(HF and phase (H), and r is the
vector cbar zyx in direct space, with lattice vectors a, b, c (defining the
periodicity of the crystal structure). The integration is performed over all vectors r in
the unit cell.
Through inverse Fourier transformation, the electron density is given by:
H
rHHHr ]2)(exp[)(1
)( iiFV
. (2.10)
The magnitude of the structure factor can be determined experimentally from X-ray
diffraction data but the phase cannot. As a consequence, the electron density cannot be
automatically determined from an X-ray diffraction data and this represents the “phase
problem” at the heart of crystal structure solution. Multiple methods now exist for
overcoming the phase problem and allowing crystal structure solution from X-ray
diffraction data. Among these are direct methods,70,71 the Patterson method,72 the
charge-flipping method73-75 and, most importantly in the context of this thesis, the direct
space strategy.76
22
Structure determination begins with the indexing of a powder diffraction pattern.
This is an essential first step and often the point at which failure occurs. The position of
peaks within the powder XRD pattern are determined and entered into an indexing
program. This program then uses an algorithm to search for relationships between the
entered peak positions, in an attempt to determine the unit cell dimensions and metric
symmetry (e.g., cubic, monoclinic, etc.). By identifying the systematic absences, along
with density considerations, the space group of the structure may then be assigned. If
additional information is required, solid-state NMR spectroscopy can be used to
determine the number of crystallographically distinct molecules within the unit cell.
After indexing, the powder pattern must undergo profile fitting using, for
example, the Le Bail method77 (as performed in this work) or the Pawley method.78 The
aim of profile fitting is to fit the complete powder XRD profile via refinement of
variables that describe the peak positions (determined by the unit cell parameters and
the zero-point shift parameter), the background intensity distribution, the peak widths,
the peak shapes and the peak intensities. The input values of the unit cell parameters are
those obtained in the indexing stage, with the refined values resulting from this fitting
representing a more accurate set of unit cell parameters. It must be emphasized that no
structural model (other than the unit cell parameters used to determine peak position) is
used in profile fitting and that the intensities of the peaks represent a set of variables
that are refined to give an optimal fit to the experimental powder XRD pattern without
reference to any structural model. The aim of profile fitting is not to determine the
crystal structure but to obtain reliable values for the previously mentioned variables that
describe the powder XRD pattern in preparation for the subsequent steps of the structure
determination process.
The goodness of fit (between the calculated and experimental powder XRD
patterns) is most commonly measured by the weighted and unweighted profile
R-factors, Rwp and Rp, respectively, defined as:
i ii
i ciii
yw
yywR
2
2
wp 100 (2.11)
i i
i cii
y
yyR 100wp (2.12)
23
where, yi is the intensity of the ith point in the experimental powder pattern, yci is the ith
point in the calculated powder pattern and wi is a weighting factor for the ith point
defined as ii yw 1 .
The peak shape depends on several instrumental and sample dependent effects.
Instrumental effects on peak broadening include the measurement geometry, the
distribution of wavelengths within the incident radiation (monochromated radiation
produces narrower peaks than non-monochromated radiation), the type of detector used
and the degree of collimation of the X-ray beam. The sample dependent effects include
broadening caused by particle size, stress and strain within the particles and disorder or
stacking faults. The overall peak shape can be defined using a number of functions, the
most comprehensive of which used in this project is a pseudo-Voigt function as
described by Thompson, Cox and Hastings79 convoluted with a correction for peak
asymmetry caused by axial divergence as described by Finger, Cox and Jephcoat.80
Axial divergence is an instrumental effect that causes asymmetric line broadening at
low angles.
After refinement of the unit cell and profile parameters in the profile-fitting
procedure, structure solution via the direct space technique is carried out. The direct
space strategy involves the creation of trial structures built from chemical knowledge.
Knowledge of the molecules present in the crystal structure allows these molecules to
be placed as fragments into the structure solution. By treating molecules as a single
fragment, rather than treating each individual atom as a freely moving entity, the
complexity of the structure solution process is greatly reduced. Each molecular
fragment can be described by a set of variables, three positional variables x, y, z, three
orientational variables and, depending on the molecule, a number of torsional
angle variables 1 2 … n. During a direct space calculation, a powder XRD pattern is
simulated for each trial structure and is compared against the experimental powder XRD
pattern by determining the Rwp value (the weighted powder R-factor). As in the case of
Le Bail fitting and Rietveld refinement, the lower the value of Rwp, the better the quality
of the fit between calculated and experimental powder XRD patterns.
The program used for direct-space structure solution in this project, EAGER,81-86
uses a genetic algorithm (GA) for exploring the Rwp hypersurface. The GA applies an
evolutionary technique by carrying out mating and mutation operations on a population
24
of trial structures. In an analogous way to evolution in the natural world, the “genetic
code” of the trial structures is the previously described positional, orientational and
torsional structural parameters. These parameters are changed by mating and mutation
operations and the “fittest” structures are defined as those with lowest Rwp. The
mutation operation will randomize a set number of parameters (the number of
parameters to be randomized is defined before beginning the calculation), e.g., a trial
structure with a set of parameters (three positional, three orientational and two torsional
angle): x, y, z, 1, 2 may mutate, randomizing four parameters to give a new
trial structure with parameters: x, Y, z, 1, 2, where the values of four
parameters are the same as the previous structure and the values of the other four are
new and random. The mating operation will mix the parameters of two parent structures
to form two daughter structures, e.g., parent trial structures with parameters:
x, y, z, 1, 2 and X, Y, Z, 1, 2 will mate to produce two daughter
structures with possible parameters: x, Y, Z, 1, 2 and X, y, z, 1, 2
(other combinations of the parameters from the parent structures are possible).
The GA is usually carried out on multiple processors in parallel, allowing multiple
independent populations. A graphical representation of the GA process is shown in
Figure 2.3. In generation M there will be a population of Np trial structures. In
generation 0, i.e., the starting point for the calculation, the initial population of trial
structures is constructed through randomization of the structural parameters, i.e., the
molecules within each trial structure will have random positions and orientations within
the unit cell and random values of torsional angles. From the population in generation
M, the mating operations are carried out, i.e., Nm pairs of structures are selected (where
Nm is the defined number of mating operations per generation) and for each pair, two
daughter structures created through the mixing of the structural parameters of the parent
structures, as described above. These daughter structures are added to the initial
population to produce an intermediate population of Np + 2Nm trial structures.
From this intermediate population, mutation operations are carried out, i.e., Nx
mutant structures (where Nx is the defined number of mutation operations) are created
from the trial structures of the intermediate population through randomization of one or
more (the number of randomized parameters is set within the input file) of the structural
parameters of the mutant structures. These mutant structures are passed along to the
next generation, i.e., generation M + 1. In addition, from the intermediate population,
25
the (Np − Nx) best trial structures (where “best” is defined as those with the lowest Rwp)
are selected and passed along to the next generation, M + 1. The structure solution is
judged to have been successful when multiple independent populations from parallel
GA calculations have reached the same, chemically sensible structure with a good fit to
the powder XRD pattern. This structure then becomes the starting point for the final
step in structure determination, Rietveld refinement.87
Rietveld refinement involves fitting the simulated powder pattern for a structure
to an experimental powder pattern. However, unlike Le Bail fitting, during Rietveld
refinement the intensities of the diffraction peaks are fully simulated from the structure.
Rietveld refinement allows the movement of each individual atom within the crystal
structure and so allows improvement upon the structure produced by the direct space
GA structure solution. Usually, however, soft restraints are imposed during the
refinement, based on chemical knowledge. Bond lengths and angles within molecules
are set to be close to known values (though still allowed to vary near to these values)
and planar parts of molecules (such as aromatic rings) are restrained to remain close to
planarity. In addition, Rietveld refinement allows the refinement of the isotropic
Figure 2.3: Schematic of the genetic algorithm (GA) process used during structure
solution with the program EAGER.
26
displacement parameters mentioned previously. In terms of displacement parameters,
constraints are set to cause all atoms within a molecule (or multiple similar molecules)
to have a common value of the isotropic displacement. The value of Uiso for hydrogen
atoms is usually fixed at a value of 1.2 times that of the atom to which it is bonded. Site
occupancy can also be refined during Rietveld refinement. This allows disorder to be
taken into account and the refinement of populations contributing to the disordered
sites.
2.3 - Solid-State Nuclear Magnetic Resonance Spectroscopy
Nuclear magnetic resonance (NMR) spectroscopy88 is a widely used and highly
informative technique for the study of the chemical environment of a target nucleus. In
the process of crystal structure determination from powder XRD data, solid-state NMR
can provide information that may be greatly beneficial (for example, identification of
the number of molecules within the asymmetric unit). It may also be used to investigate
molecular motion and disorder within crystalline structures. In addition, solid-state
NMR can be used to study amorphous and highly disordered phases, gaining
information on internuclear distances and bond angles which may not be possible
through other techniques.
In liquid-state NMR spectroscopy, fast motion of the molecules allows for
complete orientational averaging, i.e., the sample is isotropic. In solid-state NMR,
however, there is generally significantly less molecular motion (on the time scale of
NMR spectroscopy) and so the sample is anisotropic. Chemical shift depends on the
orientation of the crystal with respect to the external magnetic field. Solid-state NMR is
generally performed on powder samples containing a large number of randomly
oriented crystallites. Each crystallite orientation, due to the chemical shift anisotropy,
will have a different chemical shift and so the peaks within a solid-state NMR spectrum
become broadened. In addition, the peaks are broadened by dipolar coupling of both
heteronuclear (e.g., 1H-13C dipolar coupling) and homonuclear varieties (e.g., 1H-1H
dipolar coupling). To reduce the line widths observed in solid-state NMR, two
techniques are used: (i) magic-angle spinning and (ii) high-power decoupling (most
commonly applied to the 1H nuclei).
When performing magic-angle spinning, the solid-state NMR sample (contained
within a sample rotor) is spun about an axis forming an angle of 54.74° (the “magic
27
angle”) with respect to the applied magnetic field, B0. The anisotropic part of the spin
Hamiltonian (excluding large quadrupoles) is averaged to zero by spinning at this angle.
By calculating a value for R (the angle between the spinning axis of the rotor and the
external magnetic field) that satisfies the expression:
0)1cos3( 22
1 R , (2.13)
we find the magic-angle of approximately 54.74°. Provided that the spinning frequency
is significantly greater than the static line width caused by the anisotropy, the peak will
be narrowed.
The interactions that cause this broadening may be homogeneous or
inhomogeneous. For inhomogeneous interactions (such as chemical shift anisotropy), a
single crystal would give sharp peaks and so the broadening found in a powder sample
is due to the range of orientations present. In this case, for full line sharpening, the
spinning frequency only needs to be greater than the static line width for a single
crystallite, not greater than the static line width for the entire powder. However, in
addition to line sharpening, a set of spinning sidebands are produced. These are peaks
spaced at intervals of the spinning frequency from the centreband at isotropic chemical
shift. While these additional peaks can complicate the NMR spectrum (thus, higher
spinning frequencies where the spinning sidebands are weaker and spaced further apart
may be advantageous), additional information on the chemical shift anisotropy can be
extracted from the spinning-sideband manifold. For homogeneous interactions (such as
homonuclear 1H-1H dipolar coupling), broad lines are still observed in single crystals
due to the variety in the size and orientation of internuclear distances. In this case, very
high spinning frequencies are required to exceed the static line width of the crystallite
and produce a high-resolution spectrum (thus, solid-state 1H NMR requires very high
spinning frequencies).
Solid-state NMR spectra are often recorded using the cross-polarization89 (CP)
technique, particularly to observe spins of low natural abundance, such as 13C. Cross-
polarization involves the transference of magnetization from a nucleus of high Larmor
frequency (such as 1H) to nuclei of low Larmor frequency (X). By transferring
magnetization from 1H to X, two advantages are achieved: (i) a signal enhancement is
achieved with a maximum value determined by the ratio of the Larmor frequencies of
the 1H and X nuclei, i.e., 0(1H)/0(X) (equivalent to the ratio of the gyromagnetic
28
ratios H/X where H and X are the gyromagnetic ratios of 1H and X, respectively), and
(ii) because the magnetization originates from 1H, signal recovery by relaxation is
dependent on T1(1H) rather than T1(X). With respect to (i), the maximum signal
enhancement for a 1H→13C CP experiment compared to a 13C direct excitation
experiment is approximately fourfold [as 0(1H)/0(
13C) ≈ 4]. With respect to (ii),
values of T1(1H) are generally smaller than values of T1(X), allowing shorter recycle
delays to be used in a CP experiment than would be the case in a direct-excitation
experiment.
The pulse sequence for this technique is shown in Figure 2.4. Initially a 90° pulse
is applied to the 1H channel of the spectrometer, before applying a spin-lock pulse to the
1H channel and a contact pulse to the X channel (where X is the second nucleus in the
CP experiment). The spin-lock and contact pulses are set in such a way as to allow
transference of magnetization between 1H and X nuclei. This transference relies on
dipolar coupling between the 1H and X nuclei and, thus, cross-polarization cannot occur
in the solution state (as dipolar coupling is averaged out by motion in solution). After
Figure 2.4: Pulse sequence for the simple cross-polarization magic-angle spinning
NMR experiment.
29
the contact pulse, signal is collected for the X nuclei while performing 1H decoupling to
remove heteronuclear dipolar coupling (as this is not completely averaged by magic-
angle spinning at typical spinning frequencies). This decoupling may be performed
either by continuous irradiation of the 1H nuclei, or as used in this thesis, a sequence of
pulses such as SPINAL64.90
After being perturbed from equilibrium by a pulse, the spins of the nuclei will
revert back to their equilibrium situation in a process known as relaxation. The main
processes involved in relaxation are spin-lattice relaxation (also known as longitudinal
relaxation) and spin-spin relaxation (also known as transverse relaxation). Spin-lattice
relaxation describes the process of regaining the equilibrium of the z component of the
spin magnetization vector. Generally, for solid samples, this process is describable as a
single exponential with a relaxation time constant T1. Knowledge of T1 is essential when
performing experiments where repeated pulses are utilized (i.e., the vast majority of
solid-state experiments). Sufficient time must be allowed between each pulse (the time
between pulses is known as the pulse delay or recycle delay) for spin-lattice relaxation
to complete. If the recycle delay is too short and spin-lattice relaxation is incomplete
when the next pulse occurs, the amount of signal acquired will be reduced. However, if
the recycle delay is longer than needed, spectrometer time is wasted. To ensure that the
information gained from signal intensities is quantitative, a recycle delay of 5T1 is used
(a delay of this length results in recovery of 99.3% of the signal). If quantitative
intensity information is not required, the recycle delay can be reduced below the value
of 5T1 which, while reducing the signal gained in each repetition, allows more
repetitions to be performed within a given period of time and can therefore lead to an
increase in signal to noise (the optimum recycle delay to produce the maximum signal
to noise is 1.26T1). However, when working under compromised relaxation conditions,
one or more “dummy” acquisitions (pulses where the resultant data are not saved) may
be undertaken before acquisition of the actual data. This ensures that the diminution in
signal caused by the insufficient recycle delay occurs prior to the collection of data and,
thus, changes in relative intensity between acquired spectra remain quantitative
(provided that there are no changes in the value of T1).
Spin-spin relaxation describes the process of regaining the equilibrium of the xy
component of the spin magnetization vector (zero at equilibrium). The typical relaxation
time constant for spin-spin relaxation is given the symbol T2. In solids, T2 is generally
30
much shorter than T1 (with values of T2 for protons in a rigid solid typically on the order
of tens of microseconds). The value of T2 is related to the width of peaks in the
spectrum. A driving force for relaxation is molecular motion within the sample. Solids
containing molecules with mobile groups (e.g., methyl groups will generally undergo
facile rotation about the axis of the –CH3 bond) will have shorter values of T1 than
completely rigid molecules.
2.4 - In-Situ Solid-State NMR Studies of Crystallization
As mentioned in Chapter 1, solid-state NMR can be used to study the evolution in
solid phases present during the crystallization process. This allows the detection of
transient, unstable crystalline forms that would not be identified by ex-situ analysis of
the final crystallization product. The crystallization process is carried out within a solid-
state NMR rotor and, therefore, the crystallization solution must be prepared within the
rotor.
To prepare the crystallization solution, the solute and solvent are measured by
mass into the bottom of a 4 mm zirconia solid-state NMR rotor. The rotor is spun in a
centrifuge for a few seconds to ensure that the sample is at the bottom of the rotor. The
rotor is then sealed using a plastic insert (made of either PTFE or Kel-F®, depending on
the solvent used), as shown in Figure 2.5, to ensure that the crystallization solution does
not escape from the rotor when it is placed under magic-angle spinning. Following
sealing, the rotor is placed within the NMR spectrometer and spun under magic-angle
Figure 2.5: Schematic showing the sealed solid-state NMR rotor used to carry out in-
situ solid-state NMR studies of crystallization.
31
conditions (usually at 8 to 12 kHz). The spectrometer is tuned at the temperature that
will later be used for data acquisition. The sample is then warmed to an upper
temperature for a set period of time (usually 1 hr), to allow for complete dissolution,
before cooling to the crystallization temperature. After reaching this temperature,
acquisition of NMR spectra is begun.
The NMR spectra may either be acquired as solely solid-state measurements
(using CP, as previously described) or as interleaved solid-state and liquid-state
measurements (using direct-excitation) the so-called CLASSIC (Combined Liquid- And
Solid-State In-situ Crystallization) NMR method, developed during the work described
in Chapter 3. As stated in Section 2.3, CP cannot occur in the liquid state, so by using a
CP pulse sequence, the solid-state can be selectively observed, and the liquid state is
“invisible”. This means that NMR spectroscopy can be used to selectively monitor both
solid and liquid phases during crystallization using the same instrument, simply by
changing the pulse sequence used. During these measurements, good time resolution is
obviously critical to the success of the experiment. The time resolution is dependent on
both the number of individual scans required to produce each spectrum and the recycle
delay between the acquisition of each scan. Solution-state measurements tend to use a
short recycle delay (1 to 3 seconds) while those of solid-state measurements vary
depending on the longitudinal 1H relaxation time (T1) of the solid phase in question; a
longer relaxation time requires a longer recycle delay to gather signal during the
acquisition. Measurements of nuclei of low natural abundance (e.g., 13C) require a
greater number of scans to achieve sufficient signal to noise within each individual
spectrum. A natural-abundance 13C sample with a long relaxation time will therefore be
very challenging to achieve suitable time resolution to observe crystallization. To
improve matters, the in-situ experiments carried out during this project were performed
using the UK 850 MHz Solid-State NMR Facility located at the University of Warwick.
The high field of this spectrometer increases the signal-to-noise ratio and so allows for
greater time resolution. Spectral resolution is also increased at high field, allowing for
separation of peaks which may overlap at lower field strength.
2.5 - X-Ray Photoelectron Spectroscopy
X-ray photoelectron spectroscopy (XPS)91 involves the study of core electron
binding energies through their emission by absorption of X-ray radiation. The elemental
32
composition, oxidation state and charge within a sample can all be studied. In XPS
(Figure 2.6), a sample, held under vacuum, is illuminated with a beam of X-rays
(normally using monochromatic Al K X-rays). The X-ray photons cause excitation of
the electrons from core atomic orbitals and XPS is focused on studying the kinetic
energies of those electrons ejected from the material by this process. The kinetic energy
and intensity of the emitted electrons are measured by a detector allowing determination
of the binding energy:
)( KhB EEE , (2.14)
where EB is the binding energy of the electron, Eh is the energy of the incident X-ray
photon, EK is the kinetic energy of the emitted electron and is the work function of the
spectrometer. The electrons emitted from the sample must escape to the vacuum in
order to be detected. While the X-ray photons may penetrate deeply into the sample,
electrons emitted below the surface will not be detected. These electrons interact readily
with atoms within sample and may undergo inelastic collisions or become trapped either
through recombination with an ion or within an excited state of an atom. This causes
XPS to only be sensitive to the surface of the sample and not to the bulk, causing
surface contamination effects to be highly problematic.
Atomic binding energies are sensitive to the electronic structure of the species in
question. This means that XPS is able to distinguish, for example, between the neutral
Figure 2.6: Schematic of X-ray photoelectron spectroscopy. An incident X-ray photon
causes excitation and expulsion of a core electron causing it to be ejected from the
surface with kinetic energy Ek.
33
nitrogen within an amine –NH2 group and the charged nitrogen within an ammonium
–NH3+ group and so allow differentiation between the uncharged, non-zwitterionic and
zwitterionic forms of amino acids or between a co-crystal and a salt.
2.6 - Differential Scanning Calorimetry
Differential scanning calorimetry (DSC)92 is a technique for studying the thermal
properties of a sample. The sample (of known mass on the order of a few milligrams) is
held within a sealed aluminium pan, which is placed on a plinth within the calorimeter.
An empty reference pan is placed on a second plinth. The calorimeter then measures the
difference in heat flow into the sample pan compared to the reference pan. Measuring
this heat flow allows calculation of the heat capacity of the sample and the enthalpy and
entropy change of any phase transitions, for example a polymorphic transformation,
melting or crystallization. This arrangement of a heat flux “turret-type” differential
scanning calorimeter is shown in Figure 2.7.
Figure 2.7: Schematic of a heat flux, turret differential scanning calorimeter.
Temperatures are measured by the thermocouples indicated by black circles.
Examples of different features that may be observed in DSC data are shown in
Figure 2.8. This example shows, on heating, a glass transition (i.e., the transformation
of a glass to a supercooled liquid or rubber state), a phase transformation (e.g., the
transformation of one polymorph to another) and finally a melting peak. On cooling, a
crystallization peak is observed.
In DSC data, the peak value of thermal events is often found to vary depending on
heating/cooling rate and the sample used. As shown in Figure 2.9, it is possible to
calculate an onset temperature for the event by finding the point where the tangent of
34
Figure 2.8: Schematic demonstrating various thermal events that may be observed
using differential scanning calorimetry (DSC). Exothermic events are shown plotted
upwards.
Figure 2.9: Schematic demonstrating the calculation of the onset temperature of a
transformation in differential scanning calorimetry (DSC) data.
35
the peak crosses the interpolated baseline. The onset temperature does not generally
depend on heat/cooling rates or sample and so is a more reliable value to report.
2.7 - Thermogravimetric Analysis
Thermogravimetric analysis (TGA) measures the change in mass of a sample
upon heating. A sample is contained within an aluminium pan and its mass is measured
as it is subjected to increasing temperature. As volatile components of the sample are
lost, the mass of the sample decreases. Analysis of this decrease in mass allows the
determination of, for example, the stoichiometry of a solvate structure. From the
observed mass decrease associated with the desolvation event on heating, the number of
moles of solvent can be determined and related to the number of moles of starting
material.
An example TGA plot is shown in Figure 2.10. In this example, two partial losses
of mass are observed and may be assigned as two desolvation processes; for example, a
dihydrate transforming to a monohydrate followed by transformation to the anhydrous
form. At higher temperature, a large decrease in mass corresponding to the
decomposition of the sample is observed.
Figure 2.10: Schematic demonstrating the change in mass that may be observed in
thermogravimetric analysis (TGA) data.
36
2.8 - Dynamic Vapour Sorption
Dynamic vapour sorption (DVS) is a gravimetric method for studying the
hydration and dehydration of samples through varying humidity. A sample of known
mass is subjected to increasing followed by decreasing relative humidity (each in a
stepwise fashion) and the change in mass measured. The humidity is increased in a
stepwise fashion, with time allowed for equilibration of the sample to the new humidity
level. As in the case of TGA, observed changes in mass can offer insight on the number
of molecules of water involved in each step of hydration/dehydration processes. An
example DVS plot is shown in Figure 2.11.
DVS is widely used within the pharmaceutical93 and food science industries.94
The propensity and mechanisms for water absorption and loss from samples are
critically important when considering formulation, processing and storage of
pharmaceuticals. Hygroscopicity may be seen as an undesirable characteristic in a drug
compound as water absorption may cause “stickiness” of the sample, complicating
processing, including the pressing of tablets. Some medications, such as sodium
valproate,95 deliquesce on exposure to humidity.
Figure 2.11: Schematic demonstrating a typical dynamic vapour sorption (DVS)
measurement. Shown are two hydration/dehydration events with hysteresis, i.e., the
relative humidity value for the event is different depending on whether the humidity is
increasing or decreasing.
37
Chapter 3: Discovery of a New System Exhibiting
Abundant Polymorphism: m-Aminobenzoic Acid
3.1 - Introduction
m-Aminobenzoic acid (m-ABA) is a simple molecular organic solid, part of the
aminobenzoic acid family. All three isomers are known to be polymorphic:
o-aminobenzoic acid (o-ABA) has three known polymorphs,47,48,96 m-ABA had two
known polymorphs97 prior to the present work and p-aminobenzoic acid (p-ABA) has
two known polymorphs.52,53 m-ABA has a wide range of applications and it is used as a
fundamental organic building block in the synthesis of drugs such as analgesics and
antihypertensives.98 The literature reports the use of m-ABA in the chemical passivation
of metals including iron,99 steel99 and aluminium.100
m-ABA contains both an acidic (carboxylic acid) group and a basic (amine)
group, allowing the molecule to potentially exist in non-zwitterionic and/or zwitterionic
forms (Figure 3.1): a case of prototropic tautomerism. In solution, the tautomer of
m-ABA present is found to vary between solvents. m-ABA is known to be present as
the non-zwitterionic tautomer in DMSO,101 1,4-dioxane102 and chloroform,103 almost
entirely as the non-zwitterionic tautomer in methanol (with a proportion of non-
zwitterions greater than 98%),104 and predominantly as the zwitterionic tautomer in
water (proportion of zwitterions ranging from 50% to 78%).104,105
Prior to the present work, there were two known polymorphs of m-ABA, denoted
Form I and Form II. These two forms were identified by infrared spectroscopy in
1971,97 and the m-ABA molecules assigned as zwitterionic in Form I and non-
Figure 3.1: Molecular structure of m-ABA as (a) the non-zwitterionic tautomer and (b)
the zwitterionic tautomer.
38
zwitterionic in Form II. The crystal structure of Form II was subsequently determined
by single-crystal XRD,51 confirming that the molecules in Form II are non-zwitterionic
(with head-to-head dimers formed by hydrogen-bonding of m-ABA molecules through
the carboxylic acid group). XPS studies50 on o-ABA, p-ABA and m-ABA also
confirmed the presence of zwitterionic m-ABA molecules in Form I and non-
zwitterionic molecules in Form II. Form I has no known crystal structure, presumably
due to difficulty in growing single crystals of sufficient size and quality for single-
crystal XRD. Previous work found that, under ambient conditions, Form I is more stable
than Form II, with DSC measurements demonstrating that the relationship between the
forms is enantiotropic; Form I converts to Form II at 157.25 °C.49
The initial aim of this work was to determine the crystal structure of Form I of
m-ABA directly from powder XRD data. However, structure determination of Form I
from powder XRD data proved to be intractable due to challenges encountered in the
indexing stage of the structure determination process. The resulting discovery of
abundant polymorphism in m-ABA led to the study of the system by in-situ solid-state
13C NMR and the development of the Combined Liquid- And Solid-State In-situ
Crystallization NMR (CLASSIC NMR) method.
3.2 - Experimental Methods
m-ABA was purchased from Sigma-Aldrich (purity >98%) and powder XRD of
this sample revealed that it was a new polymorph, denoted Form III. Samples of Form I
were prepared through fast evaporation of solvent from a solution in either methanol or
ethanol, or by cooling a supersaturated solution in dimethyl sulphoxide (DMSO). Form
II was prepared by heating Form III of m-ABA until melting was observed to begin; the
sample was then removed from the heat source before melting had completed and
allowed to cool to ambient temperature. Form III was prepared by slow evaporation of a
solution in methanol (evaporation occurring overnight) or by anti-solvent crystallization
induced by addition of water to a solution of m-ABA in either DMSO or methanol.
Form IV was prepared by sublimation of m-ABA at 140 °C onto a glass “cold finger”
held at ambient temperature.
Form V was prepared by the complete melting of m-ABA to form thin films
followed by rapid cooling to ambient temperature (or lower temperature by using an ice
bath). Crystallization of bulk m-ABA produced Form II, with crystallization of Form V
39
favoured by cooling a thin film of the melt. Thin films of m-ABA melt were prepared
by melting a small amount of Form III of m-ABA in a flat bottomed “boat” constructed
out of aluminium foil. This foil boat was heated on a hotplate until complete melting
was observed, at which point the boat was removed from the heat source. Rapid cooling
was induced by placing the boat onto the surface of a water/ice mixture. Heating of
m-ABA to melting was carried out in a glove box under an atmosphere of nitrogen.
Melting m-ABA in the presence of air was found to cause the sample to rapidly change
colour (becoming purple), presumably as a consequence of the sample undergoing a
chemical reaction in the presence of oxygen.
Powder XRD measurements, both for fingerprinting purposes and for structure
determination, were carried out using a Bruker D8 diffractometer operating in
transmission mode (CuK1 radiation, germanium monochromated). Samples were
either held flat between two pieces of tape or, if preferred orientation was suspected,
packed into a glass capillary. As described later, the powder XRD patterns of Forms III,
IV and V were indexed using algorithms contained within the CRYSFIRE powder
indexing suite,106 prior to profile fitting using the Le Bail technique77 using the program
GSAS with EXPGUI.107,108 Structure solution was carried out using the direct-space
genetic algorithm technique within the program EAGER, followed by final Rietveld
refinement using the program GSAS with EXPGUI.
XPS measurements were carried out on samples of the five polymorphs of m-
ABA using a Kratos Axis Ultra DLD system with an Al K X-ray source operating at
120 W. Data were collected with pass energies of 160 eV for survey spectra and 20 eV
for high-resolution scans. The system was operated in hybrid mode, using a
combination of magnetic immersion and electrostatic lenses and data were acquired
over a surface area of ca. 300 to 700 m2. A magnetically confined charge
compensation system was used to minimize charging of the sample surface and all
spectra were acquired with a 90° take off angle. A high vacuum with base pressure
10–9 Torr was maintained during data acquisition.
Solid-state 13C NMR measurements on all five polymorphs were carried out using
a Chemagnetics Infinity Plus spectrometer (1H Larmor frequency, 300.2 MHz, 13C
Larmor frequency, 75.48 MHz; 4 mm rotor; MAS frequency 12 kHz; ramped 1H→13C
CP109 with contact time, 2 ms; TPPM 1H decoupling). A range of recycle delays were
40
required due to the differing 1H spin-lattice relaxation times (T1) of the polymorphs but,
in each case, the recycle delay was at least five times the value of T1. More information
on values of T1 is given in Section 3.3.
All DSC measurements were carried out using a TA Instruments Q100 differential
scanning calorimeter. All samples were contained within hermetically sealed aluminium
pans. A range of heating and cooling rates were used between 1 °C/min and 20 °C/min.
Solubility measurements were carried out by mixing a known mass of m-ABA (of
a specific polymorph) with a known mass of solvent and sealing within a glass sample
vial. A range of mass ratios were prepared. The samples were held at 20 °C for several
days with frequent agitation to allow for maximal dissolution. Visual inspection of each
sample was used to make a qualitative judgment of dissolution (i.e., if the sample had
dissolved or not). By carrying out further studies using narrower ranges of mass ratios
in the region corresponding to complete dissolution, the solubility of each sample was
measured.
3.3 - Discovery and Characterization of New Polymorphs of m-ABA
A sample of m-ABA (>98% purity) was purchased from Sigma-Aldrich. The
powder XRD pattern of this sample did not match that of either known form of m-ABA
and was, therefore, established to be a monophasic sample of a new polymorph of
m-ABA, hereafter designated Form III. Form III was found to be produced from slow
crystallization from methanol (the crystallization taking from ca. 18 hours to several
days). Faster crystallization from methanol, for example, induced by flowing a stream
of nitrogen gas across the surface of the solution, was found to favour crystallization of
Form I. Concomitant crystallization of Form I and Form III was commonly observed
from methanol.
DSC measurements were carried out on the sample of Form III (further details of
the thermal analysis experiments are given later). The powder XRD pattern of the
sample collected from the DSC pan after heating to 200 °C and cooling to ambient
temperature was found to differ from both the two previously known forms and the new
Form III. This phase was found to be a monophasic sample of a new polymorph and
designated Form IV. Further experiments led to the discovery that Form IV is most
reliably produced from the sublimation of m-ABA onto a surface.
41
In an attempt to replicate the conditions of the DSC measurement in the
laboratory, a fifth polymorph, Form V of m-ABA, was discovered. Form V is found to
crystallize from thin films of molten m-ABA. Crystallization of bulk molten m-ABA is
found to produce Form II and indeed, attempts to crystallize thin films of m-ABA often
result in concomitant crystallization of Forms II and V. Rapid cooling of the melt was
found to favour the crystallization of Form V.
The powder XRD patterns of all five polymorphs of m-ABA are shown in
Figure 3.2. The crystal structures of Forms III, IV and V were determined directly from
powder XRD data. Further details on the structure determination process are given in
Section 3.4. Attempts to index the powder XRD pattern of Form I were unsuccessful,
almost certainly due to the inability to accurately pick peaks for indexing in the region
24° ≤ 2 ≤ 30°, in which there is a high degree of peak overlap. Synchrotron powder
XRD data were collected on Form I m-ABA at the Diamond Light Source. However, no
Figure 3.2: Powder XRD patterns recorded for samples of the five polymorphs of
m-ABA.
42
improvement of the peak resolution in the problematic region was observed, suggesting
that the peak overlap is sample dependent and not a problem of instrument resolution.
As previously mentioned, m-ABA is known to exist in two tautomeric forms,
zwitterionic in Form I and non-zwitterionic in Form II. To establish the tautomeric form
present in each of the new polymorphs, XPS measurements were carried out
(Figure 3.3). In particular, the N(1s) region is diagnostic to the tautomeric state of the
m-ABA molecule. Forms I, III and IV have an N(1s) peak at 402 eV, while Forms II
and V have an N(1s) peak at 399.5 eV. A higher binding energy for the N(1s) electrons
suggests that the nitrogen carries a positive charge. The higher binding energy peaks for
Forms I, III and IV suggest that these forms contain a charged ammonium group and
hence the m-ABA molecules are zwitterionic, whilst the lower binding energy peaks for
Forms II and V suggest that they contain an uncharged amine group and hence the
m-ABA molecules are non-zwitterionic. These results agree with the previous
literature50 assigning the m-ABA molecules in Form I as zwitterionic and those in Form
II as non-zwitterionic. The assignments of Forms III and IV as containing the
zwitterionic tautomer and Form V as containing the non-zwitterionic tautomer were
used in the final Rietveld refinement.
Figure 3.3: N(1s) XPS spectra for the five polymorphs of m-ABA showing a clear
distinction between the zwitterionic (Form I, Form III and Form IV) and non-
zwitterionic (Form II and Form V) forms.
43
Solid-state 13C NMR data were acquired for all five polymorphs of m-ABA
(Figure 3.4). From the spectra it is possible to determine the number of molecules
within the asymmetric unit in the crystal structure. Each molecule of m-ABA contains
seven unique carbon environments and so should display, at most, seven resonances in a
13C NMR spectrum (provided there is no coupling). Forms I, II and IV all have more
than seven peaks in their solid-state 13C spectra and so must contain at least two
molecules in the asymmetric unit. Forms III and V, on the other hand, have less than
seven peaks and so show no evidence of more than one molecule in the asymmetric
unit. It should be noted that this result does not provide conclusive evidence that there
are not two or more molecules present, as multiple independent resonances may be
“accidentally” unresolved.
The 1H spin-lattice relaxation time (T1) is found to vary between the polymorphs
as follows: Form I, 0.6 s; Form II, 190 s; Form III, 1.7 s; Form IV, 0.5 s; and Form V,
7.2 s. From these values, it can be seen that the T1 values for the zwitterionic
polymorphs are significantly shorter than those of the non-zwitterionic polymorphs and,
in particular, Form II. This relaxation is driven by molecular motion within the crystal
structure and so the faster relaxation of the zwitterionic forms may be attributed to
Figure 3.4: High-resolution solid-state 13C NMR spectra of the five polymorphs of
m-ABA.
44
rotation of the ammonium group. Forms II and V instead contain an amine group which
may not be as free to rotate. The result that Form V has much shorter T1 than Form II
may be ascribed to the presence of dynamic disorder and this is further discussed in the
structural description below.
3.3.1 - Structural Description
The crystal structures of Form III, Form IV and Form V of m-ABA were
determined directly from powder XRD data and their structural parameters are given
(along with those of the previously determined Form II) in Table 3.1.
Form II Form III Form IV Form V
Space
Group P21/c P21/a P1̄ P21/a
a / Å 5.047(1) 21.3393(6) 3.80003(7) 14.7870(7)
b / Å 23.06(1) 7.29604(16) 11.55389(32) 4.95659(18)
c / Å 11.790(5) 3.77733(8) 14.6333(4) 9.0814(4)
° 90 90 110.5047(14) 90
° 105.47(3) 94.8240(12) 100.0115(35) 95.2119(21)
° 90 90 96.5930(14) 90
V / Å3 1322.45(87) 586.018(33) 595.108(35) 662.85(7)
Table 3.1: Structural parameters for the polymorphs of m-ABA with known structures
(Form II previously known;51 Form III, Form IV and Form V determined during this
work).
In the crystal structure of Form III (Figure 3.5), the asymmetric unit contains a
single molecule of m-ABA present as the zwitterionic tautomer. The structure can be
described in terms of bilayers of m-ABA molecules, with the plane of the bilayer
parallel to the (100) plane. In the internal region of a given bilayer, adjacent m-ABA
molecules are engaged in N–H···O hydrogen bonding with the aryl rings projected
either above or below the mean plane of this hydrogen-bonded region, forming the
upper and lower surfaces of the bilayer. The upper and lower surfaces of adjacent
45
bilayers (formed of the aryl rings) are in contact with each other via van der Waals
interactions. Each ammonium group is within close proximity of four carboxylate
groups of adjacent m-ABA molecules. On the basis of geometric criteria, restraints were
used to produce the most favourable hydrogen-bonding arrangement, as shown in
Figure 3.6. This arrangement results in two fairly linear hydrogen bonds to oxygen
atoms of two neighbouring molecules (N–H···O angles: 161.9°, 158.0°), with the third
hydrogen involved in a bifurcated hydrogen bond to two oxygen atoms (each oxygen
atom belonging to a different m-ABA molecule). However, it should be noted that
rotation of the ammonium group about the C–N bond allows formation of other,
geometrically less optimal sets of N–H···O hydrogen bonds. This would suggest that
the energy barrier to rotation of the ammonium group is likely relatively low. Within the
bilayer, the planes of all the aryl groups are aligned parallel to each other. In the
periodic repeat along the a-axis, there are two symmetry related bilayers (related by the
a-glide operation), which differ in the orientation of the constituent m-ABA molecules.
Figure 3.5: Crystal structure of Form III of m-ABA. Shown is the view along the c-axis,
allowing the bilayer nature of the structure to be seen. Carbon atoms are shown in grey,
hydrogen atoms in white, oxygen atoms in red and nitrogen in blue. Hydrogen bonds
are represented by green dashed lines. Red dashed lines indicate the interface between
adjacent bilayers.
46
Figure 3.6: Hydrogen-bonding interactions in the crystal structure of Form III of
m-ABA between the ammonium group and oxygen atoms of four neighbouring
molecules. All the molecules shown are in the same bilayer, illustrating that the planes
of the molecules in a given bilayer are parallel to each other.
In the crystal structure of Form IV (Figure 3.7), there are two independent
molecules of m-ABA within the asymmetric unit, both of which are present as the
zwitterionic tautomer. Although the crystal symmetry of Form IV (triclinic, space
group: P1̄) differs from Form III (monoclinic, space group: P21/a), the two structures
share several common features. The structure of Form IV, as in the case of Form III,
contains bilayers of m-ABA molecules, though with the bilayer plane parallel to the
(011̄) plane. Within the bilayer, there are two different orientations of the aryl rings
47
(from the two independent molecules of m-ABA in the structure), differing substantially
from the structure of Form III (in which there is only one orientation of the aryl rings
within the bilayer). Similar to the structure of Form III, adjacent molecules within the
bilayer are engaged in N–H···O hydrogen bonds between ammonium and carboxylate
groups, with the aryl ring again projected above and below the mean plane of the
hydrogen-bonded region (forming the upper and lower surfaces of the bilayer). Adjacent
bilayers (related by translation) are in contact via van der Waals interactions between
the aryl rings. The ammonium group of each independent molecule of m-ABA is in
close proximity to four carboxylate groups of neighbouring molecules (as observed in
Figure 3.7: Crystal structure of Form IV of m-ABA. Shown is the view along the a-axis,
allowing the bilayer nature of the structure to be seen. Hydrogen bonds are represented
by green dashed lines. Red dashed lines indicate the interface between adjacent
bilayers.
48
Form III). Again, restraints were used during structure refinement to produce the most
geometrically favourable hydrogen bonding arrangement around the ammonium group
(Figure 3.8). As in the case of Form III, rotation of the ammonium group leads to the
formation of other, less geometrically favourable hydrogen-bonding arrangements,
again suggesting that the barrier to rotation is probably low.
Figure 3.8: Hydrogen-bonding interactions in the crystal structure of Form IV of
m-ABA. (a) Shown between the ammonium group of one molecule in the asymmetric
unit and oxygen atoms of three neighbouring molecules. (b) Shown between the
ammonium group of the other molecule in the asymmetric unit and oxygen atoms of
three neighbouring molecules. In each panel, all the molecules shown are in the same
bilayer, illustrating that the planes of the molecules in a given bilayer exhibit two
distinct orientations.
The crystal structure of Form V contains a single molecule of m-ABA in the
asymmetric unit as the non-zwitterionic tautomer. The hydrogen-bonding arrangement
must, therefore, be different from those in Form III and Form IV. In Form V, the
carboxylic acid groups of the m-ABA molecules form hydrogen-bonded dimers, )8(2
2R
in graph set notation,110-112 an arrangement that cannot occur when m-ABA is present as
the zwitterionic tautomer. Detailed consideration of the Rietveld refinement, including
the inspection of difference Fourier maps, indicated that the structure of Form V in fact
49
exhibits disorder in terms of two orientations of the m-ABA molecule in the average
crystal structure. These two orientations are related by a 180° flip about an axis passing
along the HO2C–Caryl bond and, therefore, have different positions for the amine group
and the hydrogen and oxygen atoms of the carboxylic acid group but share the same
positions for the six carbon atoms of the aryl ring and three of the four hydrogen atoms
attached to the aryl ring (excluding the hydrogen atom in the meta position with respect
to both the amine and carboxylic acid groups). The refined occupancies of the two
molecular orientations are 0.602(5) and 0.398(5), and their relative positions in the
refined structure are shown in Figure 3.9.
Figure 3.9: Molecules of the two disordered components in the crystal structure of
Form V of m-ABA. Atoms unique to the higher occupancy component are shown in
cyan, atoms unique to the lower occupancy component in magenta (with the two
positions of the disordered hydrogen atom shown in green), and atoms common to both
components are shown in grey (carbon) and white (hydrogen).
It must be noted that a crystal structure from XRD data gives an averaged
representation of the actual structure taken over all unit cells present in the crystal and
averaged over time. In the case of Form V of m-ABA, the average structure has disorder
involving two molecular orientations although, from XRD data alone, the exact spatial
distribution of the two molecular orientations cannot be determined directly.
Consequently, two possible scenarios arise, both of which are compatible with a
molecule disordered over two orientations (denoted A and B with respective
orientations xA and 1 – xA): (i) a structure in which there are ordered domains containing
50
only orientation A (with such domains representing a fraction xA of the crystal) and
ordered domains containing only orientation B (with such domains representing a
fraction 1 – xA of the crystal), or (ii) a structure in which each individual molecule
throughout the crystal has a statistical probability of having orientation A (with
probability xA) or orientation B (with probability xB). In case (i), the molecular
orientations in each individual domain are ordered and the disorder in the average
structure arises from averaging over all the different domains. In case (ii), the disorder
exists at the molecular level and, in principle, the orientation of any given molecule may
be uncorrelated with the orientations of its neighbours. Consequently, in case (i), the
molecules in a given domain have a well-defined set of interactions with their
neighbours (analogous to an ordered crystal structure) whereas, in case (ii), a range of
different local intermolecular environments may exist depending on the orientation of a
given molecule and the orientations of its neighbours. In certain cases, disorder of the
type described in case (ii) may be ruled out if it transpires that an individual molecule in
orientation A surrounded by molecules in orientation B (or vice versa) would give rise
to unfavourable intermolecular interactions between neighbouring molecules.
In the case of Form V of m-ABA, intermolecular hydrogen bonds of reasonable
geometry can form in each of the following situations (in which the molecular
orientation with occupancy of 0.602 is defined as A and the molecular orientation with
occupancy of 0.398 is defined as B): (1) an ordered domain containing only molecular
orientation A, (2) an ordered domain containing only molecular orientation B, (3) a
molecule in orientation A surrounded by neighbouring molecules in orientation B, and
(4) a molecule in orientation B surrounded by molecules in orientation A. Situations (3)
and (4) represent the type of disorder described in case (ii) above. In addition, in
situations (3) and (4), there are no unfavourable interactions of the type which would
categorically rule out a disordered model of the type observed in case (ii). Thus, from
the knowledge of the average crystal structure determined from powder XRD in this
work, it cannot be established whether the disorder in the average crystal structure of
Form V of m-ABA represents disorder between two ordered domains (each containing
only one molecular orientation), statistical orientational disorder of individual
molecules, or a model of disorder that lies somewhere between these two extremes.
However, as mentioned earlier, the T1(1H) relaxation time of Form V (7.2 s) is
substantially shorter than that of Form II (190 s). The substantially faster relaxation of
Form V could be explained by the occurrence of a dynamic process in which the
51
m-ABA molecules interconvert between the two orientations over time. As T1 relaxation
is driven by molecular motion, m-ABA molecules swapping between the two
orientations would be expected to result in a shorter T1 than in the case with no dynamic
interconversion (i.e., static disorder). The coordinated interconversion of an entire
domain between the two molecular orientations would be significantly less likely than
the same interconversion occurring on a molecular level, suggesting that case (ii), as
described above, is perhaps the more likely hypothesis. In addition, the peaks in the
solid-state 13C NMR spectrum of Form V appear to be broadened. This broadening
indicates a range of chemical environments and, thus, would be most readily explained
by the molecular disorder described in case (ii) [or for the presence of very small
domains in case (i)].
To further illustrate the intermolecular hydrogen-bonding arrangement that would
arise in the types of structure described previously, an ordered domain containing only
molecular orientation A [i.e., situation (1), Figure 3.10(a)] has neighbouring molecules
linked by N–H···O hydrogen bonds, involving one N–H bond of each amine group and
the carboxyl oxygen of each carboxylic acid group. An ordered domain containing only
molecular orientation B [i.e., situation (2), Figure 3.10(b)] has pairs of molecules linked
by both intermolecular N–H···O and N–H···N hydrogen bonds. It should be noted that
there is additional disorder in terms of the position of the hydrogen atom engaged in the
N–H···N hydrogen bond due to the presence of a crystallographic inversion centre at
the midpoint of the N···N distance. Due to this inversion centre, the averaged crystal
structure has equal populations of the N–H···N and N···H–N hydrogen-bonding
arrangements linking the two nitrogen atoms. The positions of the hydrogen atoms in
the amine and carboxylic acid groups are the specific hydrogen positions considered to
give the most favourable sets of intermolecular hydrogen bonds and restraints were used
during Rietveld refinement to maintain this arrangement.
As mentioned earlier, the structure of Form V bears some similarity to the
previously reported structure of Form II. Form II of m-ABA also contains
non-zwitterionic molecules in )8(2
2R dimers. However, Form II has two independent
molecules in the asymmetric unit whereas Form V has one independent molecule. In
addition, there are differences in the hydrogen bonding of the amine group. In Form II,
N–H···N hydrogen bonds form between the two independent m-ABA molecules to
52
Figure 3.10: Crystal structure of Form V of m-ABA viewed along the b-axis (green
dashed lines indicate hydrogen bonds) for (a) ordered domains containing the
molecular orientation of higher occupancy (i.e., A) and (b) ordered domains containing
the molecular orientation of lower occupancy (i.e., B). Note that the aryl rings have the
same positions and orientations in panels a and b, but the hydrogen bonding
arrangements are different.
53
create a “zigzagged” N–H···N–H···N–H···N–H hydrogen-bonded chain parallel to the
a-axis. Formation of N–H···O hydrogen bonds also occurs between the amine groups
within this chain and the carboxylic acid dimers, such that an amine group of one
independent molecule is hydrogen bonded to two other amine groups (each belonging to
the second type of independent m-ABA molecule, acting as a hydrogen-bond donor to
one and a hydrogen-bond acceptor to the other) and to an oxygen in a carboxylic acid
group.
3.3.2 - Stability
A range of experiments and observations have led to insights concerning the
relative stabilities of the five polymorphs of m-ABA. Under dry conditions at ambient
temperature and pressure, Form II and Form V (the two non-zwitterionic polymorphs)
are observed to be unstable. Form V transforms to Form II over ca. 2 days and Form II
transforms to Form I over ca. 7 days. Form I, Form III and Form IV are found to be
stable under dry ambient conditions over several months. However, when in contact
with a saturated solution of m-ABA in methanol (i.e., as a slurry), Form I and Form IV
are found to transform over ca. 12 hours to produce Form III. In the present work, the
relative stabilities of Form I and Form IV were unable to be determined, as slurry
experiments involving these two polymorphs result in the formation of Form III. From
this set of observations, Form III is assigned as the most stable of the known
polymorphs of m-ABA under ambient conditions.
The solubility of Form III was measured at 20 °C in acetonitrile, ethyl acetate and
methanol to allow comparison with the corresponding solubility data previously
reported by Svärd et al.49 for Form I and Form II. The solubility data for these three
polymorphs are given in Table 3.2.
Solubility (mg/g)
Solvent Form I Form II Form III
Acetonitrile 7.15 ± 0.006 23.39 ± 0.05 4.65 ± 0.05
Ethyl acetate 8.78 ± 0.27 29.61 ± 0.11 4.95 ± 0.05
Methanol 51.73 ± 0.38 50.3 ± 0.4
Table 3.2: Solubilities of Form I, Form II and Form III of m-ABA in acetonitrile, ethyl
acetate and methanol, quoted as mass of solute (mg) per unit mass of solvent (g).
54
For each of the solvents studied, the solubility of Form III is lower than that of
Form I and Form II. Lower solubility among a set of polymorphs is indicative of higher
thermodynamic stability, suggesting that Form III is indeed the most stable polymorph
of m-ABA discovered thus far.
DSC measurements have previously been reported49 for Forms I and II and
indicate that, on heating, Form II undergoes only melting at 178 °C and no other
transformation. Form I, on the other hand, is reported to either undergo a transformation
to Form II at 157 °C, or undergo melting at 172 °C. Forms I and II therefore have an
enantiotropic relationship. During this work, DSC measurements were performed on the
three new polymorphs of m-ABA in order to characterize their thermal properties.
Form III (Figure 3.11) is found to undergo an endothermic phase transition at
154 °C on heating. No reverse transition is found on subsequent cooling and the product
phase can be identified by powder XRD as Form I. As Form III is the most stable
polymorph at ambient temperature and an endothermic phase transition is observed to
Figure 3.11: DSC data for Form III of m-ABA recorded with a heating/cooling rate of
20 °C min–1. Exotherms are plotted upwards.
55
produce Form I upon heating, the relationship between Form I and Form III is assigned
as enantiotropic, with Form III having higher stability below the transformation
temperature (including ambient temperature). The fact that the reverse transformation
from Form I to Form III is not observed on cooling during the DSC measurement is
likely attributable to kinetic factors.
Form IV (Figure 3.12) undergoes an endothermic phase transition at 169 °C on
heating, which is not reversible on cooling. The resultant phase following this
transformation is found by powder XRD to be Form II. Form V (Figure 3.13) is
observed to undergo melting at 167 °C. However, the melting endotherm in the DSC
thermogram is broad. In addition, there appears to be a hint of an additional peak at a
temperature just below melting but, unfortunately, this peak could not be separated from
the melting endotherm.
The density rule is a useful rule-of-thumb for judging the relative stability within
polymorphic systems, with higher density associated with higher stability (though there
are known counter examples). The densities of the four polymorphs of m-ABA with
Figure 3.12: DSC data for Form IV of m-ABA recorded with a heating/cooling rate of
10 °C min–1. Exotherms are plotted upwards.
56
known structures (noting that Form I has no known structure) are as follows:
Form II, 1.378 g cm–3; Form III, 1.553 g cm–3; Form IV, 1.522 g cm–3; and Form V,
1.372 g cm–3. Thus, the two zwitterionic forms with known structure (Form III and
Form IV) are significantly denser than the two non-zwitterionic forms (Form II and
Form V). It is noteworthy that the polymorph of m-ABA assigned experimentally as the
most stable at ambient temperature (Form III) is also the densest.
3.4 - Details of Structure Determination
3.4.1 - Structure Determination of Form III of m-ABA
The powder XRD pattern of Form III of m-ABA was indexed using the program
TREOR113 within the CRYSFIRE program suite,106 as a monoclinic cell with metric
symmetry: a = 21.38 Å, b = 7.30 Å, c = 3.78 Å, β = 94.8° (V = 587.7 Å3). From
systematic absences, the space group was assigned as P21/a and from density arguments
the asymmetric unit was determined to contain a single molecule (Z′ = 1). Le Bail fitting
was performed on the powder XRD pattern, using the program GSAS,107 achieving a
Figure 3.13: DSC data for Form V of m-ABA recorded with a heating/cooling rate of
10 °C min–1. Exotherms are plotted upwards.
57
high quality of fit [Figure 3.14(a), Rwp = 2.45%, Rp = 1.77%] and the refined unit cell
and profile parameters were used in the structure solution calculations.
The program EAGER was used to carry out structure solution with 16
independent GA calculations performed. The evolution involved 100 generations for a
population of 100 structures, with 40 mating operations and 30 mutation operations
carried out per generation. The m-ABA molecule (used in the structure solution without
hydrogen atoms) was represented by seven structural variables: three positional
variables, three orientational variables and a single torsional variable (around the
aromatic to carboxylate group bond). After 100 generations, the same good quality fit
was found in each of the 16 calculations and so was used as the starting point for
Rietveld refinement.
Rietveld refinement was carried out using the program GSAS.107 Standard
restraints were applied to bond lengths and angles and planar restraints were used to
maintain the planarity of the aromatic ring and carboxylate group. These restraints were
relaxed during the refinement procedure. Hydrogen atoms were added to the m-ABA
molecule as defined by standard geometries. Intermolecular bond-length restraints were
used to orient the ammonium group so as to give the optimum hydrogen bonding
environment. A global Uiso value was refined for all non-hydrogen atoms, with the value
Uiso for the hydrogen atoms set to be 1.2 times that of the non-hydrogen atoms. The
March-Dollase function114,115 was used to allow for preferred orientation within the
sample. The final Rietveld refinement gave a high quality fit to the powder XRD data
[Figure 3.14(b), Rwp = 2.45% and Rp = 1.80%] with the following refined parameters:
a = 21.3393(6) Å, b = 7.29604(16) Å, c = 3.77733(8) Å, β = 94.8240(12)°;
V = 586.018(33) Å3 (2θ range, 4 to 70°; 3867 profile points; 74 refined variables).
58
Figure 3.14: (a) Le Bail fit and (b) Rietveld fit for the powder XRD pattern of Form III
of m-ABA. The experimental data is shown as red crosses, the calculated fit as a green
line and the difference between the two as a magenta line. Peak positions are displayed
as black tick marks.
59
3.4.2 - Structure Determination of Form IV of m-ABA
The powder XRD pattern of Form IV of m-ABA was indexed using the program
ITO,116 again within the CRYSFIRE suite106 with a triclinic cell with metric symmetry:
a = 3.82 Å, b = 11.60 Å, c = 14.69 Å, α = 110.5°, β = 92.8°, γ = 96.6° (V = 602.5 Å3).
From density arguments, the unit cell must contain four molecules of m-ABA (Z = 4).
The space group was assigned as P1̄ with two molecules in the asymmetric unit (Z′ = 2).
Le Bail fitting was carried out on the powder XRD pattern of Form IV, giving a high
quality of fit [Figure 3.15(a), Rwp = 2.82%, Rp = 2.10%].
The program EAGER was used to carry out structure solution with 16
independent GA calculations performed. The evolution involved 100 generations for a
population of 100 structures, with 40 mating operations and 30 mutation operations
carried out per generation. Each m-ABA molecule (used in the structure solution
without hydrogen atoms) was represented by seven structural variables: three positional
variables, three orientational variables and a single torsional variable (around the
aromatic to carboxylate group bond), amounting to a total of 14 structural variables.
After 100 generations, the same good quality fit was found in each of the 16
calculations and so was used as the starting point for Rietveld refinement.
Rietveld refinement was carried out using the program GSAS.107 Standard
restraints were applied to bond lengths and angles and planar restraints were used to
maintain the planarity of the aromatic ring and carboxylate group. These restraints were
relaxed during the refinement procedure. Hydrogen atoms were added to the m-ABA
molecules as defined by standard geometries. Intermolecular bond-length restraints
were used to orient the ammonium group so as to give the optimum hydrogen bonding
environment. A global Uiso value was refined for all non-hydrogen atoms with the value
of Uiso for the hydrogen atoms set to be 1.2 times that of the non-hydrogen atoms. The
March-Dollase function114,115 was used to allow for preferred orientation within the
sample. The final Rietveld refinement gave a high quality fit to the powder XRD data
[Figure 3.15(b), Rwp = 2.86% and Rp = 2.17%] with the following refined parameters:
a = 3.80003(7) Å, b = 11.55389(32) Å, c = 14.6333(4) Å, α = 110.5047(14)°,
β = 92.7424(16)°, γ = 96.5930(14)°; V = 595.108(35) Å3 (2θ range, 4 to 70°; 3867
profile points; 125 refined variables).
60
Figure 3.15: (a) Le Bail fit and (b) Rietveld fit for the powder XRD pattern of Form IV
of m-ABA. The experimental data is shown as red crosses, the calculated fit as a green
line and the difference between the two as a magenta line. Peak positions are displayed
as black tick marks.
61
3.4.3 - Structure Determination of Form V of m-ABA
The powder XRD pattern of Form V of m-ABA was indexed using the program
TREOR,113 again within the CRYSFIRE suite106 with a monoclinic cell with metric
symmetry: a = 14.76 Å, b = 4.95 Å, c = 9.09 Å, β = 95.2° (V = 662.1 Å3). From
systematic absences, the space group was assigned as P21/a and from density arguments
the asymmetric unit was determined to contain a single molecule (Z′ = 1). Le Bail fitting
was carried out on the powder XRD pattern of Form V, giving a high quality of fit
[Figure 3.16(a), Rwp = 2.45%, Rp = 1.83%] and the refined unit cell and profile
parameters were used in the structure solution calculations.
The program EAGER was used to carry out structure solution with eight
independent GA calculations performed. The evolution involved 100 generations for a
population of 100 structures, with 40 mating operations and 30 mutation operations
carried out per generation. The m-ABA molecule (used in the structure solution without
hydrogen atoms) was represented by seven structural variables: three positional
variables, three orientational variables and a single torsional variable (around the
aromatic to carboxylic acid group bond). After 100 generations, the same good quality
fit was found in each of the eight calculations and so was used as the starting point for
Rietveld refinement.
Rietveld refinement was carried out as described for the other forms of m-ABA.
However, a small impurity of Form II of m-ABA was present within the sample and
was included as an impurity phase during Rietveld refinement. Standard restraints were
applied to bond lengths and angles and planar restraints used to maintain the planarity
of the aromatic ring and carboxylic acid group. Difference Fourier plots created during
the refinement procedure strongly suggested the presence of positional disorder of the
amine group with the two positions related by 180° flip of the molecule about the
HO2C–Caryl bond. The carboxylic acid group was also disordered over two orientations
corresponding to the same 180° flip of the molecule about the HO2C–Caryl bond (with
the two orientations of the carboxylic acid group having the same occupancies as the
two sites occupied by the amine group). The relative occupancies of the two orientations
was refined. Hydrogen atoms were added to the molecule according to standard
geometries and the orientation of the amine group was chosen such that the most
reasonable set of intermolecular hydrogen bonds was formed, based on consideration of
62
Figure 3.16: (a) Le Bail fit and (b) Rietveld fit for the powder XRD pattern of Form V
of m-ABA. The experimental data is shown as red crosses, the calculated fit as a green
line and the difference between the two as a magenta line. Peak positions of Form V are
displayed as black tick marks and those of an impurity phase of Form II of m-ABA are
displayed as red tick marks.
63
the disorder model with domains containing only the molecular orientation of higher
occupancy and domains containing only the molecular orientation of lower occupancy
[i.e., case (i) discussed above]. Disorder was also included in the positioning of the
hydrogen-bonded hydrogen within the amine group of the lower occupancy orientation
(due to the presence of a crystallographic inversion centre at the midpoint of the
hydrogen bond). A global Uiso value was refined for the non-hydrogen atoms with the
value of Uiso for the hydrogen atoms set to be 1.2 times that of the non-hydrogen atoms.
The March-Dollase function114,115 was used to allow for preferred orientation within the
sample. The final Rietveld refinement gave a high quality fit to the powder XRD data
(Figure 3.16(b), Rwp = 4.08% and Rp = 2.79%) with the following refined parameters:
a = 14.7870(7) Å, b = 4.95659(18) Å, c = 9.0814(4) Å, β = 95.2119(21)°;
V = 662.85(7) Å3 (2θ range, 4 to 70°; 3867 profile points; 114 refined variables).
3.5 - In-Situ Solid-State NMR Study of the Crystallization of m-ABA from
Methanol
The abundant polymorphism discovered within the m-ABA system made the
crystallization of m-ABA from a solvent a prime target of study using the previously
developed in-situ solid-state NMR technique42,43 (expanded upon in Chapters 1 and 2).
Through use of solid-state 13C NMR, the identity of the solid phase present during the
crystallization process can be determined, excluding any signal from the liquid phase.
This strategy allows the formation and polymorphic evolution of the solid phase to be
monitored as a factor of time during the crystallization process. The crystallization of
m-ABA from methanol was studied using this technique.
As mentioned earlier in this chapter, crystallization of m-ABA from methanol in
the laboratory can result in the formation of either Form I or Form III (or both
concomitantly). Rapid crystallization was found, in the laboratory, to favour the
crystallization of Form I, while slower crystallization was found to favour the formation
of Form III.
3.5.1 - Methodology of In-Situ Solid-State NMR Study
This in-situ NMR study was carried out using a Bruker AVANCE III NMR
spectrometer at the UK 850 MHz Solid-State NMR Facility at the University of
Warwick (13C Larmor frequency, 213.82 MHz; 4 mm HXY probe in double-resonance
64
mode). A sample of m-ABA with natural isotopic abundance was used. To prepare the
crystallization solution, Form III of m-ABA (5.5 mg) and methanol (40.8 mg) were
measured by mass into the bottom of a zirconia NMR rotor. The rotor was spun in a
centrifuge for a few seconds to ensure that the sample was at the bottom of the rotor,
before sealing with Kel-F® inserts, as described in Section 2.4. The rotor was spun
within the spectrometer under MAS conditions (12 kHz), before heating to 85 °C for
one hour to allow for complete dissolution. The rotor was then cooled to 33 °C, before
commencing the repeated acquisition of solid-state NMR spectra. Each spectrum
involved the acquisition of 256 scans, with a pulse delay of 9 s, leading to a total time to
acquire each spectrum of 38.4 mins, representing the time resolution of the in-situ study
of the crystallization process. To record the solid-state 13C spectra, ramped 1H→13C CP
was employed (CP contact time, 1.5 ms) with 1H decoupling using SPINAL-64
(nutation frequency, 91 kHz).
3.5.2 - Results and Discussion
The evolution of the solid-state data over the course of the in-situ experiment is
shown in Figure 3.17. The presence of solid is observed in the first CP-MAS spectrum,
Figure 3.17: Solid-state 1H→13C CP-MAS NMR spectra acquired during the
crystallization of m-ABA from methanol.
65
indicating that the initial crystallization of m-ABA occurred before measurement was
started (during the cooling process). Through comparison of the spectrum with those of
the five polymorphs shown in Figure 3.4, the initial product of the crystallization is
identified as Form I of m-ABA. This phase is present without change until ca. 5 hrs
after the start of the measurement. Over the next 4 hours, a change in the chemical shift
of the carboxylic acid peak is observed (Figure 3.18), along with a change in the
distribution of the peaks in the 140 to 120 ppm region. By comparing the spectra
acquired after completion of the transformation (after ca. 9 hrs) with those in Figure 3.4,
the solid sample after the transformation is identified as a monophasic sample of
Form III. Thus, Form I has undergone a transformation to Form III of m-ABA during
the crystallization process.
Thus, the initial product of crystallization from methanol in the in-situ NMR
experiment is Form I but, over time and in contact with methanol, a transformation to
the more stable Form III occurs. These results give further explanation to the
observations found during laboratory crystallization experiments. Experiments in which
Figure 3.18: Solid-state 1H→13C CP-MAS NMR spectra acquired during the
crystallization of m-ABA from methanol showing the peak due to the carboxylate
carbon of the m-ABA molecule.
66
crystallization from methanol is rapid would be expected to result in Form I of m-ABA
(as seen in the laboratory), whereas slower crystallization experiments, in which the
crystallized product is in contact with methanol for an extended length of time, would
be expected to result in Form III of m-ABA (again, mirroring the laboratory
experiments).
3.6 - In-Situ CLASSIC NMR Study of the Crystallization of m-ABA from DMSO
During the course of this study, the in-situ NMR technique for studying
crystallization processes was expanded to include solution-state NMR, leading to the
CLASSIC NMR (Combined Liquid- And Solid-State In-situ Crystallization NMR)
technique, expanded upon in Chapter 2. By monitoring both the solid-state and solution
state 13C NMR spectra over the course of the experiment, the evolution of both the
solid-state and the solution-state can be studied essentially simultaneously. Thus, the
complimentary changes that occur in the solid and liquid phases as crystallization
proceeds can be studied. From the solid-state NMR spectra, the amount and
polymorphic identity of the solid phase present is established as a function of time.
Concomitantly, the solution becomes more dilute as crystallization proceeds, resulting
in changes in solution-state speciation and the modes of molecular aggregation, which
are revealed in the liquid-state NMR spectra. As mentioned in Chapter 2, spectra are
acquired in an alternating manner by cycling solid-state CP-MAS and liquid-state
direct-excitation pulse sequences.
Before carrying out the CLASSIC NMR study of crystallization of m-ABA from
DMSO, extensive tests of the crystallization process (involving dissolution at elevated
temperature and inducing crystallization by cooling to ambient temperature) were
carried out under normal laboratory conditions. In some cases, however, the viscous,
supersaturated solution obtained on cooling to ambient temperature remained kinetically
stable and no crystallization was observed within the timescale (up to several days) of
the experiments. In such cases, crystallization could not even be induced by agitation,
shock or further cooling (to ca. –5 °C).
3.6.1 - Methodology of CLASSIC NMR Experiment
The CLASSIC NMR experiment was carried out using a Bruker AVANCE III
NMR spectrometer at the UK 850 MHz Solid-State NMR Facility at the University of
67
Warwick (13C Larmor frequency, 213.82 MHz; 4 mm HXY probe in double-resonance
mode). The crystallization solution was prepared at ambient temperature by adding
solid m-ABA Form III (24.8 mg) into the bottom of a zirconia NMR rotor followed by
DMSO (23.2 mg), added by pipette. The rotor was spun in a centrifuge for a few
seconds to ensure that the sample was at the bottom of the rotor, before sealing with the
Kel-F® inserts, as described in Section 2.4. The sample was spun under MAS (at 12
kHz), before heating to 120 °C for one hour to allow complete dissolution of the
sample, followed by cooling to 33 °C over ca. 15 mins. The CLASSIC NMR strategy
was then applied over 15 hrs. To record the solid-state 13C NMR spectra, ramped
1H→13C CP was employed (CP contact time, 1.5 ms) with 1H decoupling using
SPINAL-64 (nutation frequency, 91 kHz). To observe the liquid-state 13C NMR spectra,
the direct-excitation 13C NMR pulse sequence was used, with a single 90° 13C pulse and
no 1H decoupling. The time to acquire each spectrum was 38.4 mins for CP-MAS and
6.4 mins for 13C direct-excitation. The effective time resolution for the CLASSIC NMR
study (the time to acquire both a solid-state and liquid-state spectrum) was therefore
44.8 mins. A correction was applied (by another member of the research group) to the
peak positions in the liquid-state 13C spectra to account for a general drift in chemical
shift caused by the cooling of the shims within the spectrometer during measurement.
3.6.2 - Results
The evolution of the solid-state 13C NMR spectra, over the course of the
experiment, is shown in Figure 3.19. The first solid-state signal emerged ca. 2 hours
after commencing the experiment, signifying the start of crystallization. From
comparison of the 13C chemical shifts with those in Figure 3.4 (by summing all the
solid-state 13C NMR spectra acquired over the experiment as shown in Figure 3.20), this
solid phase is assigned as Form I of m-ABA. The intensity of the peaks in the solid-state
spectrum then increased monotonically with time (Figure 3.21), indicating that the
amount of solid increased and, therefore, that the crystallization process continued until
levelling off at ca. 8 hrs. No further changes occur in the solid-state spectra for the
remainder of the experiment, i.e., no polymorphic transformation occurs and the final
product of the crystallization is Form I of m-ABA. This observation is in agreement
with the results determined in the laboratory.
68
Figure 3.19: Solid-state 1H→13C CP-MAS component of the CLASSIC 13C NMR data
acquired during the crystallization of m-ABA from DMSO.
The liquid-state 13C NMR spectra contain seven sharp peaks for each of the seven
13C environments in the m-ABA molecule, with only J-coupling to directly attached
protons resolved. The summed integral over the peaks in each liquid-state spectrum
allows the proportion of m-ABA in solution to be measured as a function of time
(Figure 3.21). This proportion is found to remain level until that point at which the first
Figure 3.20: Sum of all solid-state 1H→13C CP-MAS NMR spectra acquired during the
crystallization of m-ABA from DMSO using CLASSIC NMR.
69
solid signal is observed in the solid-state 13C NMR spectra (i.e., the point at which
crystallization has begun). Thereafter, the summed integral decreased with time, before
reaching a constant value at ca. 8 hrs. That the concentration of m-ABA in solution is
not observed to decrease until Form I is observed in the solid-state spectra rules out the
possibility of crystallization of Form II prior to Form I. Due to the long T1 relaxation
time of Form II of m-ABA, it would not be observable in the CP-MAS 13C spectra (T1
of 190 s compared to a pulse delay of 9 s). Crystallization of Form II prior to that of
Form I would, however, be manifest in a decrease in the total integral of the liquid-
NMR spectra.
The results from the solid-state and liquid-state measurements are, therefore, in
good agreement on the start point and end point of the crystallization process.
Figure 3.21: Fraction of m-ABA molecules in the liquid (blue) and solid (red) phases as
a function of time, established from the integrals of the liquid-state and solid-state
components of the CLASSIC 13C NMR data, respectively. To determine the fraction of
molecules in the liquid state, the integrals of the liquid-state NMR spectra were
normalized relative to the integral of the first spectrum recorded (at this stage of the
experiment, it is known that all m-ABA molecules are in the liquid state). To determine
the fraction of molecules in the solid state, the integrals of the solid-state NMR spectra
were normalized by setting the fraction ( eq
SF ) of m-ABA molecules in the solid state at
the end of the experiment as eq
L
eq
S FF 1 , where eq
LF is the fraction of m-ABA
molecules in the liquid state at the end of the experiment.
70
Crystallization began 2 hrs after measurement was started with crystal growth
continuing for the next 6 hrs. After these 8 hours, no further evolution in the total
amounts of m-ABA in the liquid and solution was observed suggesting that, by this
stage, the system comprised an equilibrium saturated solution.
A more detailed interpretation of the changes in the liquid-state 13C NMR
spectrum during crystallization can now be developed. Figure 3.22 shows the 13C
chemical shift i(t) for each site (i) in m-ABA as a function of time, relative to the
corresponding initial value istart. Initially, the system is a supersaturated solution (with
concentration ca. 1.4 times the solubility of Form III of m-ABA at 33 °C). It should be
noted that the solubility of Form I is higher than that of Form III at ambient
temperature. This fact may account for any discrepancy between the total decrease in
the integral of the liquid-state 13C NMR spectrum during the crystallization process
(Figure 3.21) and the total amount of solute that is predicted to crystallize based on the
Figure 3.22: Evolution of 13C chemical shifts in the liquid-state component of the
CLASSIC NMR data, for crystallization of m-ABA from DMSO.
71
degree of supersaturation, with respect to Form III, of the initial crystallization solution.
After crystallization begins, the supersaturation decreases with time. By the end of the
crystallization process, the system is an equilibrium saturated solution with chemical
shifts denoted ieq(DMSO). To rationalize the changes that occur in the solution state as
crystallization proceeds, it is instructive to consider the values of
iclassic = i
start – ieq(DMSO), representing the difference in chemical shift between
the initial solution (with maximum supersaturation) and the final solution (at
equilibrium saturation). Values of iclassic determined from the data in Figure 3.22 are
given in Table 3.3.
As an independent comparison, the liquid-state 13C NMR spectra for saturated
solutions of m-ABA (Form III) in water, methanol, 1,4-dioxane and DMSO were
recorded at 33 °C (recalling101-105 that m-ABA is predominantly zwitterionic in water,
almost entirely non-zwitterionic in methanol and non-zwitterionic in 1,4-dioxane and
DMSO). For each 13C site i in m-ABA, the difference in chemical shift in a given
solvent relevant to DMSO is denoted isolvent = i
eq(solvent) – ieq(DMSO), where
i
classic iwater [a] i
methanol [a] i1,4-dioxane [a]
C3 –0.27 –11.48 –0.27 –0.13
C5 –0.03 0.34 –0.30 –0.32
C1 –0.02 3.47 0.01 –0.34
C7 0.06 3.65 1.07 –1.10
C2 0.14 5.09 0.94 0.06
C4 0.16 4.44 0.84 –0.43
C6 0.25 7.27 1.19 0.39
r2 [b] 0.88 0.71 0.05
Table 3.3: Values of iclassic measured from the liquid-state component of the CLASSIC
13C NMR data for crystallization of m-ABA from DMSO and values of isolvent
determined from ex-situ liquid-state 13C NMR data. [a] For saturated solutions in water,
methanol, 1,4-dioxane and DMSO. [b] Correlation coefficients (r2) between values of
iclassic and values of i
solvent for each solvent.
72
ieq(solvent) is the chemical shift in the saturated solution at 33 °C and i
eq(DMSO)
is the chemical shift in the saturated DMSO solution at 33 °C. Comparison (Table 3.3)
of values of iclassic and i
solvent gives a very high correlation coefficient for water
(0.88), a moderately high correlation coefficient for methanol (0.71) and a very low
correlation coefficient for 1,4-dioxane (0.05). Recalling that the definitions of
iclassic and i
solvent have the same reference point [i.e., chemical shifts ieq(DMSO) for
a saturated solution of m-ABA in DMSO], the high correlation coefficients for water
and methanol suggest that the speciation and intermolecular environment of m-ABA
molecules in the supersaturated DMSO solution at the start of the CLASSIC NMR
experiment (reflected in the values of istart and hence in the values of i
classic) may
bear some resemblance to those of m-ABA molecules in saturated solutions in water
and methanol [reflected in the values of ieq(solvent) and hence in the values of
isolvent]. Based on the high correlation coefficients observed between values of
iclassic and values of i
water and imethanol, two hypotheses are now developed
concerning the nature of the supersaturated solution of m-ABA in DMSO at the start of
the CLASSIC NMR experiment.
First, the key difference (reflected in the values of iwater) between saturated
solutions of m-ABA in water and DMSO is that m-ABA is predominantly zwitterionic
in saturated water solution but non-zwitterionic in saturated DMSO solution. Thus, one
hypothesis is that, at sufficiently high supersaturation of m-ABA in DMSO, the
proportion of m-ABA molecules present as zwitterions is significantly higher than that
in an equilibrium saturated solution of m-ABA in DMSO. The fact that the magnitudes
|iclassic| are less than |i
water| suggests that the proportion of zwitterionic m-ABA
molecules in the supersaturated solution at the start of crystallization from DMSO is
less than that in the saturated solution of m-ABA in water (50 to 78% zwitterionic
tautomer in water104,105).
Second, a key difference (reflected in the values of imethanol) between saturated
solutions of m-ABA in methanol and DMSO is the presence of a hydrogen-bond donor
in methanol (while DMSO can only act as a hydrogen-bond acceptor). In saturated
methanol solution, m-ABA is almost entirely non-zwitterionic and the amine group is
likely to act as an acceptor in O–H···N hydrogen bonds. Thus, a second hypothesis is
that, in the supersaturated solution of m-ABA in DMSO at the start of the crystallization
experiment, there is an increase in the proportion of non-zwitterionic m-ABA molecules
73
present in aggregates containing intermolecular hydrogen bonds with the amine group
acting as the acceptor. As DMSO has no hydrogen-bond donors, the amine group of
non-zwitterionic m-ABA can be a hydrogen-bond acceptor only with amine groups or
carboxylic acid groups of other m-ABA molecules as the donor.
Thus, the changes (iclassic) in 13C chemical shifts observed in the crystallization
process are consistent with the supersaturated solution of m-ABA in DMSO at the start
of crystallization having a higher proportion of zwitterionic m-ABA molecules and/or a
higher proportion of non-zwitterionic m-ABA molecules present in hydrogen-bonded
aggregates, relative to a saturated solution of m-ABA in DMSO. Both scenarios
represent an increased degree of protonation of the amine group, leading to increased
positive charge on the nitrogen atom, increasing the shielding of C3 and promoting the
specific changes in 13C chemical shifts observed.
Finally, it must be noted that the evolution of the 13C chemical shift for the solvent
DMSO as a function of time in the liquid-state component of the CLASSIC NMR data
(Figure 3.22) is a direct consequence of the decrease in the concentration of m-ABA
molecules and hence in the extent to which DMSO molecules are engaged as the
acceptor in O–H···O or N–H···O hydrogen bonds with m-ABA molecules.
The crystal structure of Form I of m-ABA has not yet been determined, although
the N(1s) XPS studies carried out in the present work (Figure 3.3) indicate that the
m-ABA molecules are zwitterionic. Thus, it is expected that the ammonium and
carboxylate groups will interact via intermolecular O–···H–N+ hydrogen bonds in the
crystal structure, as these are the only hydrogen-bond donor and acceptor groups
present. From the analysis of the liquid-state component of the CLASSIC NMR data
discussed above, it can be proposed that the supersaturated solution of m-ABA in
DMSO at the start of the crystallization process contains (i) zwitterionic m-ABA
molecules, which may or may not be present as aggregates containing
O–···H–N+ hydrogen bonds (although such aggregates must ultimately be formed at
some stage on the crystallization pathway) and/or (ii) aggregates of non-zwitterionic
m-ABA molecules linked by O–H···N hydrogen bonds.
Clearly, either situation is a plausible precursor to the O–···H–N+ hydrogen bonds
that are expected to exist between m-ABA zwitterions in the crystal structure of Form I.
Although the current work cannot distinguish whether situation (i) or situation (ii) is
74
predominant, the results give clear insight into the nature of the speciation and
interactions that exist in the supersaturated pre-nucleation solution of m-ABA in DMSO
prior to the crystallization, relative to those in the saturated solution at the end of the
crystallization process. While this rationalization of the evolution of the solution-state
13C chemical shifts during crystallization are based on empirical observations of the
differences in 13C chemical shifts from m-ABA in different solvents in which the
speciation of the m-ABA molecules is known independently, it should be noted that an
alternative approach could be to compute the 13C chemical shifts via DFT calculations
on proposed solution-state clusters. Such ab initio calculations are now widely used in
the field of NMR crystallography.117-119
3.7 - Conclusions
The discovery of three new polymorphs of m-ABA during this work has
demonstrated the abundant polymorphism within this system. The fact that one of the
three new polymorphs of m-ABA discovered in the present work (Form III) is more
stable under ambient conditions than the two observed previously (Form I and Form II)
emphasizes that the earliest discovered polymorphs of a given molecule do not
necessarily include the most stable polymorph. Indeed, by extrapolating this argument,
it cannot be stated that Form III is definitely the thermodynamically stable polymorph
under ambient conditions, only that it has the highest relative stability among the
polymorphs of m-ABA discovered so far. An interesting aspect of the polymorphism of
m-ABA is the fact that the molecule is zwitterionic in some polymorphs and non-
zwitterionic in others. Among the polymorphs of o-ABA and p-ABA, the zwitterionic
tautomer appears less common than in the case of m-ABA. Thus with the exception of
Form I of o-ABA (which contains both zwitterionic and non-zwitterionic molecules in a
1:1 ratio), all other polymorphs of o-ABA and p-ABA contain only non-zwitterionic
molecules. Although none of the polymorphs of m-ABA contain both zwitterionic and
non-zwitterionic molecules within the same crystal structure, the discovery in the
current work of two new polymorphs of m-ABA containing zwitterionic molecules
(Form III and Form IV) and one new polymorph containing non-zwitterionic molecules
(Form V) suggests that the dividing line between these tautomeric forms is readily
crossed in this system.
75
Use of in-situ solid-state NMR has demonstrated that during crystallization of
m-ABA from methanol, Form I of m-ABA is formed initially, followed at a later point
by transformation into the more stable Form III. The in-situ result therefore gives
insight into the crystallization behaviour of m-ABA from methanol in a laboratory
setting: faster crystallization results in Form I while slower crystallization results in
Form III.
The CLASSIC NMR experiment extends the scope and capability of in-situ
monitoring of crystallization processes, particularly as it provides complementary
information on the time-evolution of both the solid phase and the liquid phase during
crystallization. The CLASSIC NMR strategy, applied to the crystallization of m-ABA
from DMSO, has given clear insights into the changes in the nature of speciation and
interactions as the crystallization process occurs. In spite of the simplicity of the
concept of “interleaving” the measurement of liquid-state and solid-state NMR spectra,
it is surprising that this strategy has not been exploited more widely in other scientific
fields, although a notable case is the SedNMR technique, which has been developed to
monitor the kinetics of sedimentation and fibrillization processes in biological
systems.120 Another recent development to study heterogeneous biomaterials is the
CMP121 technique. However, in contrast to CLASSIC NMR, the CMP technique
requires new and specialized probe design and has not been applied to systems
containing free liquid nor to study time-dependent processes. Within the study of
crystallization processes, the CLASSIC NMR technique reported here is unique as an
in-situ NMR strategy for simultaneously mapping the time-evolution of both the liquid
and solution phases. It is anticipated that the advantages of the CLASSIC NMR strategy
will yield significant new insights on a wide range of other crystallization systems in the
future.
76
Chapter 4: Discovery of New Solid Forms of
Triphenylphosphine Oxide and Methyldiphenyl-
phosphine Oxide by In-Situ Solid-State NMR
4.1 - Introduction
Triphenylphosphine oxide [TPPO, Figure 4.1(a)] is a very commonly encountered
chemical. TPPO is regularly used as a ligand in metal coordination122 and is a common
by-product of several organic synthesis reactions including the Mitsunobu,123
Staudinger124 and Wittig125,126 reactions. In addition, cocrystals of TPPO have been
found to have favourable crystal growth and handling conditions,56 allowing TPPO to
be used as a crystallization aid. Nevertheless, questions remain over the polymorphism
of TPPO. There are four reported polymorphs of TPPO (Figure 4.2): monoclinic I,57
orthorhombic,58 monoclinic II59 and monoclinic III.127 All four polymorphs have known
crystal structures. Of these, only monoclinic I and orthorhombic have known and
replicable methods of crystallization. Monoclinic II is reported59 to have crystallized
from a solution containing a metal complex. However, no further details on the solution,
metal complex or crystallization conditions are given. No details are known of the
method used to crystallize monoclinic III; the structure was submitted to the CSD as a
private communication. The crystal structures of three polymorphs of TPPO
hemihydrate128-130 and the structure of a diethyl ether solvate131 have also been reported.
TPPO is reported to crystallize concomitantly as the monoclinic I and
orthorhombic polymorphs by evaporation of a solution in hexane.57 The highly
polymorphic nature of TPPO, combined with the ability to crystallize as two
polymorphs concomitantly from certain solvents, made TPPO an interesting system to
study crystallization via in-situ NMR. Previous solid-state 31P NMR performed on
Figure 4.1: Molecular structures of (a) triphenylphosphine oxide (TPPO) and (b)
methyldiphenylphosphine oxide (MDPPO).
77
Figure 4.2: Crystal structures of the four polymorphs of TPPO: (a) monoclinic I, (b)
orthorhombic, (c) monoclinic II and (d) monoclinic III.
78
TPPO assigned chemical shifts of 26.8 ppm for the monoclinic I polymorph and 28.5
ppm for the orthorhombic polymorph.132
Prior to this work, the related compound, methyldiphenylphosphine oxide
(MDPPO) was not known to be polymorphic. There is a single reported crystal structure
of MDPPO (Figure 4.3).60 MDPPO finds use in a variety of organic syntheses61-63 and,
as in the case of TPPO, as a ligand. The fact that MDPPO has only one known crystal
structure while the very closely related compound TPPO is highly polymorphic made
the crystallization of MDPPO a prime target for study via in-situ solid-state NMR.
4.2 - Experimental Methods
TPPO was purchased from Sigma-Aldrich (purity >98%). This material was
found by powder XRD to be a monophasic sample of the monoclinic I polymorph of
TPPO. A sample of orthorhombic TPPO was prepared by heating monoclinic I to
melting, before cooling to ambient temperature.
Initial in-situ solid-state 31P NMR studies on the crystallization of TPPO from
solution were carried out using a 300 MHz Chemagnetics Infinity Plus spectrometer
(31P Larmor frequency, 121.5 MHz; 4 mm rotor; MAS frequency 8 kHz; ramped
1H→31P CP109 with contact time, 1.5 ms; TPPM 1H decoupling). Crystallization of
TPPO from ethanol/hexane (1:1 by mass, 31.9 mg TPPO in 14.3 mg ethanol and
Figure 4.3: The reported crystal structure of MDPPO, viewed along the a-axis. The
crystal structure is disordered with one phenyl ring occupying three possible
orientations. For clarity, only a single orientation of this phenyl ring is shown.
79
14.7 mg hexane) and ethanol/cyclohexane (1:3 by mass, 19.5 mg TPPO in 6.7 mg
ethanol and 21.4 mg cyclohexane) by mass were investigated. In each case, 16 scans
were acquired for each spectrum using a pulse delay of 120 s, resulting in each spectrum
taking 32 mins to record (representing the maximum time resolution).
A CLASSIC NMR study of crystallization of TPPO from solution in
ethanol/cyclohexane, 1:3 by mass, and acquisition of solid-state 31P NMR spectra for
powder samples of the monoclinic I and orthorhombic polymorphs of TPPO were
carried out using a Bruker AVANCE III spectrometer at the UK 850 MHz Solid-State
NMR Facility at the University of Warwick (31P Larmor frequency, 344.17 MHz; 4 mm
HXY probe in double-resonance mode). A sample of TPPO with natural isotopic
abundances was used. The monoclinic I polymorph of TPPO (15.0 mg), ethanol
(6.1 mg) and cyclohexane (19.7 mg) were measured by mass into the bottom of a
zirconia NMR rotor. The rotor was spun in a centrifuge for a few seconds to ensure the
sample was at the bottom, before sealing with PTFE inserts, as described in Section 2.4.
The rotor was spun within the spectrometer under MAS conditions (12 kHz), before
heating to 75 °C for one hour to allow for complete dissolution of the sample. The rotor
was then cooled to 20 °C, before application of the CLASSIC NMR procedure over
ca. 22 hrs. In this case, the procedure was modified such that the sequence of
acquisition was: direct-excitation liquid-state 31P NMR spectrum, direct-excitation
liquid-state 1H NMR spectrum, solid-state 1H→31P CP-MAS spectrum. Each solid-state
spectrum involved the acquisition of 16 scans, with a pulse delay of 120 s and each
liquid-state spectrum (for each of 31P and 1H) involved 16 scans with a pulse delay of 3
s, leading to a total time to acquire each spectrum of 33.6 mins, representing the
maximum time resolution of the experiment. To record the solid-state spectra, ramped
1H→31P CP was employed (CP contact time, 1.5 ms) with 1H decoupling using
SPINAL-64 (nutation frequency, 91 kHz).
MDPPO was purchased from Sigma-Aldrich (purity > 98%). The sample was
found by powder XRD to be the known crystalline phase of MDPPO. The in-situ solid-
state NMR study of the crystallization of MDPPO from toluene was carried out using a
Bruker AVANCE III spectrometer at the UK 850 MHz Solid-State NMR Facility at the
University of Warwick (31P Larmor frequency, 344.17 MHz; 4 mm HXY probe in
double-resonance mode). MDPPO (5.1 mg, natural isotopic abundance) and toluene
(34.9 mg) were measured by mass into the bottom of a zirconia NMR rotor. The rotor
80
was spun in a centrifuge for a few seconds to ensure the sample was at the bottom,
before sealing with Kel-F® inserts, as described in Section 2.4. The rotor was spun
within the spectrometer under MAS conditions (12 kHz), before heating to 80 °C for
one hour to allow for complete dissolution of the sample. The rotor was then cooled to
20 °C, before acquisition of repeated solid-state 31P NMR spectra. Each spectrum
involved the acquisition of 16 scans, with a pulse delay of 9 s, leading to a total time to
acquire each spectrum of 38.4 mins, representing the maximum time resolution of the
experiment. To record the solid-state spectra, ramped 1H→31P CP was employed (CP
contact time, 1.5 ms) with 1H decoupling using SPINAL-64 (nutation frequency,
91 kHz).
Powder XRD measurements on TPPO and MDPPO were performed using a
Bruker D8 diffractometer operating in transmission mode (CuK1). Fingerprinting of
solid samples was carried out with the sample held between two pieces of tape (foil-
type sample holder).
4.2.1 - Crystallization within a Glass Capillary
Crystallization of TPPO from ethanol/cyclohexane (1:3 by mass) was carried out
within glass capillaries to allow powder XRD data to be recorded for the crystallization
products without removal from the mother liquor (with a schematic representation of
this process shown in Figure 4.4). TPPO (ca. 0.305 g), ethanol (ca. 0.125 g) and
cyclohexane (ca. 0.385 g) were measured into a glass, liquid-state NMR tube of 5 mm
external diameter. The tube was sealed using a plastic NMR tube cap and Nescofilm®
before being submerged (leaving only the top 1/3 uncovered) in silicone oil and heated
to 60 °C. The tube was held at this temperature with frequent agitation until all solid
TPPO present was observed to dissolve in the mixed solvent. Five to six glass
capillaries of external diameter 0.7 mm were cut to a length (from sealed bottom to open
top) of ca. 15 mm. The NMR tube containing the solution was unsealed and the cut
glass capillaries were added such that the open tops of the capillaries were pointing
upwards. The NMR tube was resealed and maintained at 60 °C while vigorously
agitating the solution until the glass capillaries were observed to have filled with
solution and sunk to the bottom of the tube. The tube was unsealed, the solvent emptied
and the glass capillaries recovered as quickly as possible. The exterior of each glass
capillary was cleaned with a small amount of ethanol/cyclohexane (1:3 by mass)
81
absorbed into a facial tissue, before sealing the open end of each capillary with either
quick setting epoxy resin or superglue. Once the resin/glue had finished setting, powder
XRD data were recorded for the capillaries containing crystalline material.
4.3 - Results
4.3.1 - Triphenylphosphine Oxide
Crystallization experiments were carried out on TPPO in the laboratory. Pure
orthorhombic TPPO was found to be reliably produced by heating monoclinic I to the
melting point, allowing partial melting to occur and cooling to ambient temperature.
Crystallization of TPPO from ethanol via evaporation of the solvent was found to
produce the monoclinic I polymorph of TPPO. It is reported57 that slow evaporation of a
solution of TPPO in hexane results in the concomitant crystallization of both the
monoclinic I and orthorhombic polymorphs of TPPO. Crystallization of TPPO from
cyclohexane via evaporation was found to produce only orthorhombic TPPO.
Figure 4.4: Schematic showing the method used to carry out crystallization within glass
capillaries.
82
The solubility of TPPO is considerably higher in ethanol than in hexane or
cyclohexane. In an attempt to find a solvent mixture giving rise to concomitant
crystallization of the monoclinic I and orthorhombic polymorphs and in which TPPO
has a high solubility (allowing the system to be studied by in-situ NMR), the mixed
solvent systems ethanol/hexane (1:1 by mass) and ethanol/cyclohexane (1:3 by mass)
were tested. Evaporation of solutions of TPPO in both of these mixed solvents was
found to result in concomitant crystallization of monoclinic I and orthorhombic TPPO.
The solid-state 31P NMR spectra for the monoclinic I and orthorhombic
polymorphs of TPPO are shown in Figure 4.5. Both monoclinic I and orthorhombic
contain one molecule of TPPO in the asymmetric unit and so the NMR spectrum of
each polymorph contains a single 31P peak (at 26.4 ppm for monoclinic I and 28.5 ppm
for orthorhombic). The most intense peak in each spectrum is not, in fact, the isotropic
peak but the first spinning sideband to higher chemical shift.
Initial in-situ solid-state 31P NMR studies on the crystallization of TPPO were
performed using a 300 MHz solid-state NMR spectrometer as described in the
experimental section. Crystallization of TPPO was carried out from mixed
Figure 4.5: Solid-state 1H→31P CP-MAS NMR spectra for (a) the monoclinic I
polymorph of TPPO and (b) the orthorhombic polymorph of TPPO.
83
ethanol/hexane (both 1:1 and 1:3 by mass) and from ethanol/cyclohexane (1:3 by mass).
The signal-to-noise ratio observed during these experiments was insufficient to allow
for suitable time resolution to study changes in the solid phase over time. However, by
summation of all spectra acquired in each experiment, the solid phases present during
the crystallization process can be identified. As shown in Figures 4.6 and 4.7, no phases
are present that represent either the monoclinic I or orthorhombic polymorphs of TPPO.
Powder XRD conducted on the products of the crystallization in these two experiments
demonstrated the presence of both monoclinic I and orthorhombic phases of TPPO. The
unknown phases must, therefore, undergo a transformation to form the known phases.
Thus, the unknown phases appear to be unstable upon removal from the NMR rotor.
To increase the signal-to-noise ratio and to gain both time resolution and spectral
resolution, the in-situ solid-state 31P NMR study of the crystallization of TPPO from
ethanol/cyclohexane (1:3 by mass) was repeated at the UK 850 MHz solid-state NMR
facility. In this experiment, the CLASSIC NMR method was applied, allowing
acquisition of three sets of NMR spectra (liquid-state 31P NMR spectra, liquid-state 1H
NMR spectra and solid-state 31P NMR spectra), with the aim of monitoring the
Figure 4.6: Sum of the in-situ solid-state 1H→31P CP-MAS NMR spectra acquired
during the crystallization of TPPO from ethanol/hexane (1:1 by mass). The positions of
the peaks due to the monoclinic I and orthorhombic polymorphs of TPPO and the
position of the unknown intermediate phase are marked.
84
evolution in both the solid-state and liquid-state over the course of the crystallization
process.
The evolution in the solid-state 31P NMR spectrum over time is shown in
Figure 4.8. A signal is observed in the first recorded solid-state spectrum and, thus,
crystallization must have begun before or during the recording of the first spectrum. A
single peak is observed at 63.30 ppm (phase A). Two additional peaks are observed at
61.58 and 61.34 ppm (phase B and phase C, respectively) and grow over time
(appearing at the intensity contour level used in the plot shown in Figure 4.8 at ca.
6 hrs). The intensity of the peak due to phase A is found to decrease monotonically with
time until ca. 17.5 hrs, while the intensity of the peaks due to phase B and phase C are
found to increase over the same 17.5 hrs. After ca. 17.5 hrs, no further signal is
observed for phase A. Phase A has therefore completely transformed into phase B and
phase C. The sum of the solid-state 31P spectra recorded over the last ca. 6.1 hrs of the
experiment is shown in Figure 4.9, clearly demonstrating the existence of the two
separate peaks belonging to phase B and phase C.
Figure 4.7: Sum of the in-situ solid-state 1H→31P CP-MAS NMR spectra acquired
during the crystallization of TPPO from ethanol/cyclohexane (1:3 by mass). The
positions of the peaks due to the monoclinic I and orthorhombic polymorphs of TPPO
and the position of the unknown intermediate phase are marked.
85
The liquid-state 31P NMR spectrum was found to contain a single 31P resonance
corresponding to TPPO dissolved in the mixed solvent. The integrated intensity of the
peak due to phase A, along with the total integrated intensity of the peaks belonging to
phase B and phase C (which are positioned too close to accurately calculate separate
integrated intensities) and the integrated intensity of the liquid-phase peak are shown in
Figure 4.10. The intensities are normalized such that the maximum value in each case is
set to 1. The intensity of the peak for phase A is found to decrease from the beginning
of the experiment until ca. 17.5 hrs, indicating that the amount of this phase decreases
with time until it is completely lost at ca. 17.5 hrs. Concurrently, the intensities of the
solid-state peaks for phase B and phase C are found to increase over the same period,
indicating an increase in the amounts of these two phases. In addition, the intensity of
the liquid-state peak decreases throughout the experiment. Thus, over the course of the
Figure 4.8: In-situ solid-state 1H→31P CP-MAS NMR spectra acquired during the
crystallization of TPPO from a mixed ethanol/cyclohexane (1:3 by mass) solvent. The
region corresponding to the first spinning sidebands to higher chemical shift of the
isotropic peaks is shown. Dashed lines indicate the positions expected for peaks due to
the orthorhombic polymorph of TPPO (O), monoclinic I polymorph of TPPO (MI) and
the positions of the three unknown phases (A, B and C).
86
experiment, there is both a transformation of phase A to phase B and phase C as well as
a continuation of crystallization from solution.
Unfortunately, analysis of the changes in the chemical shifts of the peaks in the
liquid-state 1H NMR spectrum (in a similar manner to that performed for the liquid-state
13C chemical shifts during the CLASSIC NMR study of m-ABA in Chapter 3) could not
be performed due to changes in the shimming over the course of the experiment. When
the sample is heated in the spectrometer to allow dissolution, the shims (which are used
to ensure magnetic field homogeneity over the sample) are also heated and subsequently
cool over the course of the experiment. This cooling causes a change in shimming
resulting in changes in peak shape and, therefore, changes in the peak position
calculated by the peak picking procedure, thus, masking changes in chemical shift
caused by changes in solute concentration.
It must be noted that the peaks in the solid-state spectra here assigned to phases A,
B and C, do not match the peak positions belonging to either monoclinic I or
orthorhombic TPPO. It could be argued that a small error in referencing could allow the
peak assigned as phase A to match the peak due to orthorhombic TPPO and/or would
allow the peaks assigned to phase B or phase C to match the peak due to monoclinic I.
However, the difference in chemical shift between orthorhombic and monoclinic I (1.70
ppm) does not match the difference between phase A and phase B (1.72 ppm) or the
Figure 4.9: Sum of the last 11 solid-state 1H→31P CP-MAS NMR spectra (last ca.
6.1 hrs) acquired during the crystallization process,
87
difference between phase A and phase C (1.96 ppm). Therefore, even if one of these
three unknown phases is, in actuality, monoclinic I or orthorhombic TPPO, there remain
two other unknown phases.
In addition, the behaviour of phase A in this in-situ study does not appear to
match that expected for orthorhombic TPPO. Due to the common concomitant
crystallization of orthorhombic and monoclinic I polymorphs of TPPO, it would be
expected that the two polymorphs are close in relative stability. Thus, orthorhombic
TPPO would not be expected to undergo the fairly rapid transformation that phase A is
observed to undergo during this experiment. Due to the inability to reproduce the
crystallization of the monoclinic II or monoclinic III polymorphs of TPPO (and
therefore to record solid-state 31P NMR spectra for well-characterized powder samples
of these polymorphs), it cannot be ruled out that these polymorphs are present during
this in-situ experiment. Nevertheless, the results from this CLASSIC NMR study give
evidence for a possible three new solid forms of TPPO. Solvates of TPPO are obviously
Figure 4.10: Evolution in the integrated intensities of the TPPO peak in the liquid-state 31P NMR spectra (black), the peak at 63.3 ppm in the solid-state 31P spectra (red) and
the integrated intensity of the combination of the peaks at 61.58 and 61.34 ppm in the
solid-state 31P spectra (green). The intensities have been normalized such that the
maximum value in each case is set to 1.
88
also a possible identity for the phases observed here; however, so far, no ethanol solvate
or cyclohexane solvate of TPPO have been reported.
In an attempt to replicate the conditions of the in-situ crystallization and record
powder XRD data on the new phases observed, crystallization of TPPO from
ethanol/cyclohexane (1:3 by mass) was carried out within glass capillaries, using the
method described in Section 4.2.1. The majority of attempts were found by XRD to
produce only monoclinic I and orthorhombic polymorphs. However, two capillaries in a
single crystallization batch were found to contain a polycrystalline sample that was
revealed by powder XRD to match none of the known polymorphs of TPPO (Figure
4.11).
This new powder XRD pattern was indexed successfully (using the program
KOHL133 within the CRYSFIRE program suite106) with orthorhombic metric symmetry
and unit cell dimensions: a = 13.09 Å, b = 22.66 Å, c = 12.08 Å, V = 3583 Å3. Le Bail
fitting (using the program GSAS107) was also successful (using space group P222,
Figure 4.12), resulting in a unit cell of dimensions: a = 13.09427(30) Å, b = 22.6553(5)
Å, c = 12.0816(4) Å, V = 3584.04(22) Å3 (Rwp = 1.82%, Rp = 1.29%; 2θ range, 4 to 70°;
3867 profile points). However, further crystal structure determination directly from this
powder pattern was unsuccessful, presumably due to a high degree of preferred
orientation introduced by carrying out the crystallization process within the glass
capillary. Density arguments suggest that this phase may possibly be a solvate of TPPO.
The densities of the polymorphs of TPPO range from 1.22 g cm–3 (monoclinic II) to
1.24 g cm–3 (monoclinic I). Assuming Z = 8 and assuming that the new phase contains
only TPPO, the density would be 1.03 g cm–3, significantly lower than of any of the
known polymorphs of TPPO. A structure containing TPPO and ethanol in a 1:1 ratio, on
the other hand, would have density 1.20 g cm–3, much closer to those of the known
polymorphs. For comparison, the density of the diethyl ether solvate of TPPO is
1.13 g cm–3 and the density of TPPO hemihydrate ranges from 1.25 g cm–3 to
1.29 g cm–3.
89
Figure 4.11: (a) Experimental powder XRD pattern of TPPO crystallized from a mixed
ethanol/cyclohexane (1:1 by mass) solvent within glass capillaries. Simulated powder
XRD patterns of (b) monoclinic I phase of TPPO, (c) orthorhombic phase of TPPO, (d)
monoclinic II phase of TPPO and (e) monoclinic III phase of TPPO. The positions of
the first four peaks in the experimental powder XRD pattern of the crystallized sample
are shown by dashed lines.
90
Figure 4.12: Le Bail fit for the powder XRD pattern of the new phase of TPPO
crystallized from ethanol/cyclohexane (1:3 by mass) within a glass capillary. The
experimental data are shown as red crosses, the calculated fit as a green line and the
difference between the two as a magenta line. Peak positions are displayed as black tick
marks.
3.4.2 - Methyldiphenylphosphine Oxide
Laboratory experiments were performed to investigate the crystallization of
MDPPO from toluene, both through evaporation of the solvent and through cooling a
saturated solution. Powder XRD confirmed that the crystallization products of these
experiments are monophasic samples of the known crystalline phase of MDPPO.
The solid-state 31P NMR spectrum recorded for the single known form of MDPPO
has a single 31P resonance at 30.78 ppm (Figure 4.13), corresponding to the single
molecule of MDPPO in the asymmetric unit. As shown in Figure 4.13, the most intense
peak in the spectrum is not the isotropic peak, but is the first order spinning sideband at
higher chemical shift (65.57 ppm). Small peaks from impurity phases also appear at
33.84 ppm and 29.54 ppm (though whether these peaks represent impurity compounds
or other crystalline phases of MDPPO is unknown). The T1(1H) of MDPPO is found to
be less than 1 s. The evolution in the solid-state 31P NMR spectra over the course of the
crystallization of MDPPO from toluene is shown in Figure 4.14, focused on the first-
91
order spinning sideband to higher chemical shift (as to it has the highest intensity). It is
clear from Figure 4.14 that crystallization must have begun before or during the
recording of the first spectrum. All spectra acquired throughout the crystallization
process are dominated by the intense peak at 65.57 ppm, corresponding to the known
phase of MDPPO. Significantly, however, two additional peaks of considerably lower
intensity are also observed in the early stages of crystallization, with first-order spinning
sidebands at 63.7 ppm and 64.5 ppm (corresponding to isotropic peaks at 28.8 ppm and
29.6 ppm, respectively).
Only one crystalline phase of MDPPO was known prior to this work. Thus, these
two additional peaks must represent new solid forms of MDPPO produced at a very
early stage of the crystallization process. The intensity of the peaks due to the new
forms is highest in the first acquired spectrum, decreasing as a function of time. These
two peaks are assigned as arising from two different solid phases, rather than a single
solid phase with two distinct 31P environments from the fact that the evolution of the
two peaks is distinctly different, with significantly different rates of decay of peak
Figure 4.13: Solid-state 1H→31P CP-MAS NMR spectrum for the known phase of
MDPPO. The isotropic peak is located at 30.78 ppm. Two small peaks due to impurity
phases are present at 33.84 ppm and 29.54 ppm.
92
intensity. Thus, the peak at 63.7 ppm becomes indiscernible from the background after
ca. 16 mins, while the peak at 64.5 ppm remains visible for ca. 6 hrs. From these in-situ
31P solid-state NMR results, the possibility that these unknown phases represent toluene
solvates of MDPPO cannot be ruled out. However, it must be noted that no toluene
solvate of MDPPO has been reported previously.
The two new phases of MDPPO observed in the in-situ NMR experiment may be
difficult to isolate ex situ. It is obvious that these phases are produced in small amounts
during the early stages of the crystallization of MDPPPO from toluene. However, due to
the overwhelming presence of the known phase of MDPPO, isolation of the new phases
in these proportions would very likely present an insurmountable problem. Other
crystallization procedures must therefore be researched in order to produce these two
unknown forms.
Figure 4.14: In-situ solid-state 1H→31P CP-MAS NMR spectra acquired during the
crystallization of MDPPO from toluene. The region corresponding to the first spinning
sidebands to higher chemical shift of the isotropic peaks is shown.
93
4.4 - Conclusions
Through use of in-situ solid-state 31P NMR, up to three new solid phases of TPPO
have been discovered during the crystallization of TPPO from an ethanol/cyclohexane
mixed solvent (1:3 by mass). If all three phases correspond to new polymorphs of
TPPO, the total number of polymorphs for this compound would be raised to seven
(only very rare cases such as ROY3 have more known polymorphs). Crystallization of
TPPO from ethanol/cyclohexane (1:3 by mass) within glass capillaries revealed a new
crystalline phase by powder XRD. Structure determination of this phase was
unsuccessful and work is continuing in an attempt to isolate and characterize this phase,
though density arguments suggest that it may, in fact, be a solvate of TPPO.
The in-situ solid-state 31P NMR study of the crystallization of MDPPO from
toluene has demonstrated the existence of two unstable, transient phases during the
crystallization process. These phases represent possible new polymorphs of MDPPO.
Laboratory isolation of these phases is expected to be highly difficult under the
conditions used in the in-situ NMR crystallization experiment due to the low proportion
of the unknown phases relative to the known crystalline phase of MDPPO. Other
crystallization conditions (such as a variation on solvent or temperature schedule used),
will likely need to be explored but the conditions described here represent a starting
point for further experimentation.
Both of these studies exemplify the ability of in-situ solid-state NMR studies of
crystallization to discover previously unknown, transient phases formed during
crystallization processes. In particular, the case of TPPO has demonstrated that
extrapolation of the conditions used in the in-situ NMR crystallization experiment to the
laboratory allows the characterization of previously unknown crystalline phases via
powder XRD.
94
Chapter 5: New Insights into the Preparation of Form
II of rac-Ibuprofen
5.1 - Introduction
As mentioned in Chapter 1, polymorphism is both widespread and of great
importance within the pharmaceutical industry. Common pharmaceuticals, including
aspirin16-20 and paracetamol,13-15 are known to be polymorphic and undiscovered
polymorphism was disastrous in the case of the antiretroviral ritonavir.7
Ibuprofen (Figure 5.1) is listed by the World Health Organization (WHO) as one
of the three essential non-opioid and non-steroidal anti-inflammatory medicines
(NSAIMs).134 Only the S-(+)-enantiomer of ibuprofen is biologically active.135 In vivo,
however, the R-(–)-enantiomer is converted to the S-(+)-enantiomer135 and so ibuprofen
is usually supplied as the racemate. rac-Ibuprofen melt can easily be supercooled and is
reported to undergo a glass transition at –55 °C.22 At ambient temperature, the solubility
of ibuprofen in water is fairly poor while the aqueous solubilities of the sodium and
lysinium salts of ibuprofen are much higher.136 These salts are therefore marketed for
fast-acting pain treatment and migraine relief as the higher solubility allows for faster
absorption of the drug in vivo.137
Figure 5.1: Molecular structure of ibuprofen.
The first crystal structure for rac-ibuprofen was solved21 in 1974. It was not until
2008 that polymorphism was discovered in rac-ibuprofen.22 Experiments performed on
rac-ibuprofen melt demonstrated that a new polymorph, Form II (the previously
determined structure is named Form I), could be prepared by undertaking a particular
cooling regime. It was reported that, in order to produce Form II, it was essential to
quench cool ibuprofen melt to a temperature at least 60 K below the glass transition and
to hold for an hour before warming to 258 K, at which point crystal growth of Form II
would occur. It was reported that nucleation of Form II was dependent on the initial
95
quench and mechanisms based on the formation of cracks within the glass were
postulated. The melting point of Form II was reported as 290 K, substantially (59 K)
lower than that of Form I, 349 K.
The process of quench cooling of a melt to below a glass transition to achieve
crystallization replicates that observed in other systems such as m-toluidine,138
o-benzylphenol139 and 3,3’-dimethoxy-4,4’-bis(2,2-diphenylvinyl)biphenyl.140 Similar
to the case reported for Form II of rac-ibuprofen, the quench cooling of m-toluidine is
the only known method for production of the -polymorph.
The crystal structure64 of Form II of rac-ibuprofen [Figure 5.2(b)] was determined
directly from powder XRD data recorded using synchrotron radiation. The structure of
Form II contains dimers of ibuprofen hydrogen bonded through the carboxylic acid
groups, as in the case of Form I [Figure 5.2(a)]. In Form I, the hydrogen bonding occurs
perpendicular to the chains of dimers within the structure and links ibuprofen molecules
from two different chains. In Form II, however, the hydrogen bonding links ibuprofen
molecules from within the same chain.
To explore this method of crystallization further, in the present work, both DSC
and in-situ powder XRD were utilized to give insight on whether quenching to a
temperature below the glass transition is truly necessary to the formation of Form II.
5.2 - Experimental Methods
Ibuprofen was purchased commercially as the racemic sodium salt from Fluka
(purity 99.5%). This sample was converted into the free acid by dissolution in water
followed by acidification with hydrochloric acid, leading to the precipitation of rac-
ibuprofen (due to the low solubility of the free acid in water). Extraction of rac-
ibuprofen was performed using chloroform, followed by washing with water to remove
any remaining sodium salt. rac-Ibuprofen was finally recrystallized from chloroform by
evaporation of the solvent.
To confirm the purity of the sample prepared by this procedure, solution-state 1H
NMR, gas chromatography, melting point analysis and powder XRD were performed.
The solution-state 1H NMR demonstrated the absence of any chloroform-soluble
compound other than rac-ibuprofen, with gas chromatography confirming a purity of
greater than 99%. Powder XRD showed that the only crystalline phase present was rac-
96
ibuprofen Form I and melting point analysis determined a melting onset temperature of
347.15(10) K and melting enthalpy of 25.96(45) kJ mol–1, tallying well with reported
literature values (Table 5.1).
Figure 5.2: Crystal structures of (a) Form I of rac-ibuprofen, and (b) Form II of
rac-ibuprofen, each viewed along the 21 screw axis.
97
Source Onset Melting Point (K) Melting Enthalpy (kJ mol–1)
This work 347.15 ± 0.20 25.96 ± 0.90a
Chow et al.141 344.8 ± 0.2 27.2 ± 0.4a
Seton et al.142 347.84 23.33a
Xu et al.143 348.35 ± 0.11 25.04 ± 0.10;a 26.65b
Dudognon et al.22 – 25.8 ± 0.5a
Table 5.1: Onset melting point and melting enthalpy (a determined using DSC, b determined using adiabatic calorimetry) for Form I of rac-ibuprofen.
All DSC measurements were performed using a TA Instruments Q100 differential
scanning calorimeter. Samples were contained within hermetically sealed aluminium
pans. A heating rate of 10 K min–1 and cooling rate of 20 K min–1 were used in all DSC
experiments. Each DSC experiment was performed with the same general procedure.
Initially, the sample was heated to 373 K to allow for melting of the sample, followed
by cooling to a temperature lower than the glass transition (termed the quenching
temperature). The sample was held at this quench temperature for a certain length of
time before heating to a second temperature (the annealing temperature) and again
holding for a certain length of time. Finally, the sample was heated again to 373 K, to
allow observation of the melting of any crystalline sample present followed by cooling
to room temperature. In some cases, the quenching step was removed and the molten
sample was instead cooled directly to the annealing temperature. A variety of quenching
and annealing temperatures were used, with full details given in Section 5.3.
In-situ powder XRD was carried out using a Bruker D8 diffractometer in
reflection geometry (Cu K1 radiation, germanium monochromated). An Oxford
Cryostreams Phenix was used to control the temperature of the sample. The sample of
rac-ibuprofen (of approximate mass 11 mg) was contained in an indentation on a copper
disk held within a silver-coated copper sample cup. The sample of rac-ibuprofen was
melted ex situ by heating the sample contained on the copper disk to 373 K using a
bench-top hot plate, before transferring the disk to the diffractometer (with the sample
cooling to ambient temperature during the transfer). At this point, the cryostat chamber
was evacuated and cooled at a maximum rate of approximately 2 K min–1. At each
temperature to be studied, powder XRD patterns were repeatedly recorded over time. As
mentioned in Chapter 1, the time-resolution of an in-situ study is limited in this case by
the time to record a single powder XRD pattern. The total time to record a powder XRD
pattern in this study was either 44 or 88 minutes, with these times, therefore,
98
representing the maximum time resolution. Powder XRD data were recorded over the
angle range 4° ≤ 2 ≤ 43° (step size 0.016°). The data recorded with 2 below 10° are
not shown within the figures of this chapter as they are dominated by a high background
scattering caused by the shielding of the cryostat chamber.
5.3 - Results and Discussion
5.3.1 - DSC Studies of the Crystallization of Form II of rac-Ibuprofen
As previously mentioned, the intention of this work was to explore the
crystallization conditions for Form II of rac-ibuprofen in order to discover if the initial
quench cool below the glass transition point is essential to its formation. To this end,
DSC experiments constituting a range of conditions were performed, both including and
excluding the initial low temperature quench. The full set of experiments and results in
each case is summarized in Table 5.2, with a graphical representation of the temperature
scheme used in the first experiment (a repeat of that conducted by Dudognon et al.22)
shown in Figure 5.3. The results are given as percentages of the sample crystallized as
Form II and a qualitative measurement for the presence of Form I.
Quench Anneal Result
T / K Time / h T / K Time / h Form II Form I
A 143 1 258 3 70% Yes
B 143 0 258 3 100% Yes
C 143 0 258 3 <5% No
D 193 0 258 3 <5% Yes
E 258 6 35% Yes
F 258 3 0 Yes
G 258 3 <5% Yes
H 258 1 0 No
I 258 1 <5% No
J 243 3 0 Yes
K 243 3 0 Yes
L 273 12 0 No
M 273 3 0 No
N 273 3 0 No
Table 5.2: Summary of DSC experiments conducted on rac-ibuprofen.
99
Figure 5.3: Graphical representation of the procedure used in DSC Experiment A,
repeating the experiment performed by Dudognon et al.22
In each case, the crystalline content of Form II in the sample is measured through
comparison of the melting enthalpy of the Form II peak detected during the final
heating procedure with the value of the Form II melting enthalpy reported by Dudognon
et al.22 A similar approach to find the amount of Form I present during the final heating
cannot be performed, as the melting endotherm of Form I is found to overlap with the
exotherm caused by the crystallization of Form I. The presence of Form I can therefore
only be determined in a qualitative manner, i.e., either a melting endotherm for Form I
is observed or it is not observed.
The series of experiments was completed in order, beginning with experiment A,
a repeat of the method described by Dudognon et al.,22 i.e., a quench to 143 K for one
hour, followed by an annealing for three hours at 258 K. The DSC thermogram for this
experiment is shown in Figure 5.4. As shown in Table 5.2, this method resulted in
crystallization of Form II (with approximately 70% of the sample crystallizing in this
form). Form I was also observed during the final heating. It should be noted that, during
100
the DSC experiments involving quench cooling of the sample to 143 K, a spike in heat
flow was observed in the DSC thermogram. This spike seems to correlate with the
occurrence of cracking within the glassy sample described by Dudognon et al.22 (and is
marked in Figure 5.4 as a cracking event). During some attempts at cooling below the
glass transition, some of the spikes in heat flow were followed by a loss in heat flow
signal from the sample, which was found to be caused by the sealed DSC sample pan
becoming dislodged from the plinth within the calorimeter. This observation
demonstrates that the cracking events described by Dudognon et al.22 and observed here
are occasionally violent enough to cause a substantial movement of the sample pan.
To determine if crystallization of Form II is dependent on remaining at the
quenching temperature for one hour, two experiments (B and C) were performed in
which the rac-ibuprofen melt was cooled to the quenching temperature of 193 K
followed by immediate heating to the annealing temperature (at which it was annealed
for three hours). Form II was observed to have crystallized during both of these
experiments. In B, the entire sample was estimated to have crystallized as Form II while
Figure 5.4: DSC thermogram for experiment A. For clarity, the initial heating of the
sample to 373 K and final cooling to 313 K have been excluded. Indicated are the
quenching and annealing temperatures and the thermal events occurring during the
DSC experiment, including a cracking event further expanded upon in the text.
101
in C, less than 5% of the sample was estimated to have crystallized as Form II (Figure
5.5). Additionally, in B, a melting endotherm for Form I was observed during the final
heating while this was missing in C. These two experiments, while showing
inconsistency, indicate that it is possible to crystallize Form II without holding the
sample for an extended length of time at the quenching temperature.
In experiment D, the sample was quenched at 193 K (i.e., 31 K below the glass
transition), a quenching temperature higher than that reported22 to be essential for
nucleation of Form II. During the final heating run, an endotherm was observed for both
Forms II and I. However, this endotherm corresponded to only a small amount of
crystallized Form II (less than 5%). Nevertheless, this result demonstrates that
crystallization of Form II can occur without quenching to a temperature more than 60 K
below the glass transition.
Following this result, the entire quenching step was removed from the procedure.
In experiment E, the sample of molten rac-ibuprofen was cooled directly to the
Figure 5.5: DSC thermogram for experiment C. For clarity, the initial heating of the
sample to 373 K and final cooling to 313 K have been excluded. The region of 270 to
310 K has been expanded to demonstrate the very small endotherm observed for the
melting of Form II, in this case.
102
annealing temperature (258 K) and maintained at this temperature for six hours. During
the final heating run, endotherms for both Form II and Form I were observed (with
roughly 35% of the sample crystallizing as Form II). This experiment conclusively
demonstrates that quenching to low temperature is not essential for nucleation of Form
II of rac-ibuprofen. The annealing step was then shortened to three hours (experiments
F and G) and further to one hour (experiments H and I). A small amount of Form II
(less than 5% of the total sample) was observed in one of the two experiments
conducted at each annealing time (experiments G and I), indicating that the
crystallization of Form II during these annealing times is somewhat inconsistent.
Further experiments were carried out to explore the effect of the annealing
temperature on the crystallization of Form II in the absence of the initial low
temperature quenching step. Using an annealing temperature of 243 K and annealing
time of three hours (experiments J and K), no melting endotherm was observed for
Form II during the final heating, while Form I was observed in both cases. Using an
annealing temperature of 273 K and annealing times of three hours (experiments M and
N) and a longer annealing time of twelve hours (experiment L), no evidence for the
crystallization of Form II or Form I was found.
5.3.2 - In-Situ Powder XRD Studies of the Crystallization of Form II of rac-Ibuprofen
To further investigate the crystallization of Form II of rac-ibuprofen with no
quenching step, in-situ powder XRD experiments were carried out. For comparison with
the experimental powder patterns recorded during these experiments, the simulated
powder XRD patterns of the two polymorphs of rac-ibuprofen are shown in Figure 5.6.
As described in Section 5.2, Form I of rac-ibuprofen was heated until melting had
completed, ex situ, before transfer to a laboratory powder X-ray diffractometer in
reflection mode (during which time the sample cools to ambient temperature), where it
was further cooled, under vacuum, to the desired annealing temperature.
Three in-situ powder XRD experiments were performed. In the first, the sample of
liquid rac-ibuprofen was cooled to 258 K and held at this temperature for 43.0 hours
while recording repeated powder XRD patterns. After this time, the sample was warmed
to 273 K and held for a further 23.4 hours, again recording repeated powder XRD
patterns. For the first 9.5 hours after reaching 258 K, no diffraction peaks were
observed, indicating the absence of crystalline phases within the sample. At 9.5 hours,
103
peaks corresponding to the diffraction pattern of rac-ibuprofen Form II were observed.
Subsequent measurements from 9.5 to 16.1 hours showed only a slight increase in the
intensity of these peaks. Thereafter, there were no significant changes in the peak
intensities throughout the remaining time at 258 K and Form II remained as the only
crystalline phase present during this time. After 43.0 hours at 258 K, the sample was
warmed to 273 K. After 2.9 hours at this new temperature, peaks attributed to the
presence of Form I of rac-ibuprofen were observed to appear. These peaks were found
to grow in intensity over time, with a subsequent decrease in the intensity of the peaks
due to Form II. The disappearance of the peaks due to Form II was complete by
Figure 5.6: Simulated powder XRD patterns for (a) Form I of rac-ibuprofen and (b)
Form II of rac-ibuprofen, over the range 10° ≤ 2 ≤ 40° (Cu K1 radiation) using the
published crystal structures. Marked with asterisks are distinctive peaks indicative of
the form in question (red for Form I, green for Form II).
104
8.7 hours after reaching 273 K, indicating that the transformation of Form II to Form I
had completed. No further change in crystalline form was observed at 273 K. A
schematic representation of this experiment is given in Figure 5.7 and selected in-situ
powder XRD patterns are shown in Figure 5.8. The changes in the relative amounts of
Form I and Form II were measured through integration of isolated peaks belonging to
the two forms (the region 18.8° to 20.0° for Form I and the region 20.0° to 22.0° for
Form II) and the results are displayed in Figure 5.9. The integrals during this
measurement were normalized so that the maximum value for each polymorph was set
to unity. The results from this experiment corroborate the findings of the DSC
experiments, i.e., that crystallization of Form II can occur at 258 K with no prior low-
temperature quenching step.
In the second in-situ powder XRD experiment, the sample of liquid rac-ibuprofen
was cooled directly to 273 K and held at this temperature for 46.7 hours while recording
repeated powder XRD patterns. Similar to the previous experiment, no diffraction peaks
corresponding to crystalline phases were observed for the first 14.6 hours after cooling.
After this time, however, peaks corresponding to Form II were observed, increasing in
intensity until 17.5 hours. From 19.0 hours at 273 K, peaks corresponding to Form I
were first observed, and then increased in intensity with a decrease in the intensity of
Figure 5.7: Schematic representation of the phases found in the an in-situ powder XRD
experiment studying the crystallization of rac-ibuprofen from the supercooled liquid,
initially at 258 K, followed by 273 K. Shown in grey are the patterns where no
crystalline phases are observed, blue where only Form II is observed, yellow where
only Form I is observed and green where both Form I and Form II are observed.
Marked with arrows and labelled a-d are the specific powder XRD patterns shown in
Figure 5.8.
105
Figure 5.8: In-situ powder XRD patterns collected during the crystallization of
rac-ibuprofen from the supercooled liquid, initially at 258 K (for a period of 43.0
hours), and then at 273 K (for a period of 23.4 hours). Within the entire experiment, the
specific time at which each powder XRD pattern (a-d) was recorded is shown in Figure
5.7 (labelled a-d).
106
Figure 5.9: Integrated peak intensities for the phases observed during the
crystallization of rac-ibuprofen from the supercooled liquid, initially at 258 K, followed
by 273 K. Form II is shown in green (intensity integrated over the 2 range:
20.0° – 22.0°) and Form I in red (intensity integrated over the 2 range: 18.8° – 22.0°).
The integrals were normalized so that the maximum value of each polymorph is set to
unity.
the peaks due to Form II. By 27.7 hours, the peaks due to Form II disappeared
completely. No further change in crystalline form was detected for the remainder of the
experiment. A schematic representation of this experiment along with extracted in-situ
powder XRD patterns is shown in Figure 5.10. As in the previous experiment,
integrated peak intensities for isolated distinctive peaks (again, the region 18.8° to 20.0°
for Form I and the region 20.0° to 22.0° for Form II) were calculated for the two forms
during this measurement and are displayed in Figure 5.11. This experiment shows that
crystallization of Form II can occur at 273 K. The fact that Form II was not observed in
the DSC experiments involving annealing at 273 K may be attributed to the
crystallization event being preceded by a significant lag time.
A third in-situ experiment was carried out, cooling the sample of rac-ibuprofen
melt to 243 K, in order to assess whether, in a similar manner to the second experiment,
crystallization of Form II occurs at this temperature after a suitably long time. The
sample was held at 243 K for a total of 93 hours. However, no crystalline phases were
observed to form at any stages of this experiment.
107
Figure 5.10: In-situ powder XRD patterns collected during the crystallization of
rac-ibuprofen from the supercooled liquid at 273 K. Shown are extracted powder XRD
patterns recorded during the timeframe 13.1 to 29.2 hours. Shown on the right is a
representation of the complete experiment, showing scans where no crystalline phases
are observed in grey, scans where Form II is observed in blue, scans where Form I is
observed in yellow and scans where both are observed in green.
Figure 5.11: Integrated peak intensities for the phases observed during the
crystallization of rac-ibuprofen from the supercooled liquid at 273 K. Form II is shown
in green (intensity integrated over the 2 range: 20.0° – 22.0°) and Form I in red
(intensity integrated over the 2 range: 18.8° – 22.0°). The integrals were normalized
so that the maximum value of each polymorph is set to unity.
108
The DSC and in-situ powder XRD experiments, taken together, demonstrate that
the low temperature quench described by Dudognon et al.22 is not essential to the
formation of Form II. The crystallization of Form II is observed at a maximum
temperature of 258 K in the DSC experiments and 273 K in the in-situ powder XRD
experiments. This difference may be accounted for in the long lag time observed before
crystallization at 273 K during the in-situ measurements (14.6 hours). This is longer
than the time held at 273 K during the DSC experiments (12 hours) and so it is possible
that if held at 273 K for long enough in the DSC measurement, the crystallization of
Form II may be observed. This would be complicated, unfortunately, by the
transformation of Form II to Form I at 273 K. Additionally, the DSC results show some
irreproducibility, with repeated experiments under the same conditions producing
varying crystallization results, as is often the case in polymorphic systems. This
demonstrates that a high degree of care must be taken when investigating crystallization
from a super-cooled melt.
5.4 - Conclusions
Although pharmaceutical applications of Form II of rac-ibuprofen are necessarily
limited by the fact that its melting point is below ambient temperature, the results
reported here are nevertheless significant within the broader context of understanding
the phase behaviour and polymorphism in the rac-ibuprofen system. Indeed, a
significant observation from this work concerns the kinetic stability of supercooled
liquid rac-ibuprofen, which had previously not been reported to undergo a
transformation to a crystalline phase within the temperature range to which
pharmaceuticals are typically subjected. In contrast, this work has demonstrated that
maintaining supercooled liquid rac-ibuprofen at 273 K for less than a day is sufficient
to lead to crystallization of Form II (followed, after a further period of time, by the
formation of Form I via a solid-state polymorphic transformation). Importantly, this
work has shown that low-temperature quenching of amorphous rac-ibuprofen is not a
prerequisite for the crystallization of Form II. In more general terms, the observations
reported here may also have relevance to other systems for which low-temperature
quenching of supercooled liquid phases is used as a strategy to produce specific
crystalline polymorphs (such as m-toluidine138).
109
Chapter 6: Expanding the Solid-State Landscape of
L-Phenylalanine: Discovery of Polymorphism,
New Hydrate Phases and Rationalization of
Hydration/Dehydration Processes
6.1 - Introduction
Phenylalanine (Phe), shown in Figure 6.1, is one of the 20 directly genetically-
encoded amino acids (those encoded directly in DNA by a codon). It is also one of the
essential amino acids for humans, i.e., humans are unable to biosynthesize
phenylalanine. As in the case of all amino acids other than glycine, Phe is chiral and
exists as both an L- and a D-enantiomer.
Figure 6.1: Molecular structure of L-Phe.
Phenylalanine is used industrially in the production of the artificial sweetener
aspartame (Figure 6.2), the methyl ester of the dipeptide of L-aspartic acid and
L-phenylalanine. Both an anhydrous solid form65 and a hydrated solid form66 of L-Phe
have been reported. The crystal structure of enantiomerically pure Phe is known65 (with
further information given below), while, prior to this work, there was no known crystal
structure for any L-Phe hydrate phase. During the production of L-Phe for use in
aspartame synthesis, it is highly desirable to crystallize L-Phe as the anhydrous form,66
due to the better separation and handling properties when compared to the hydrate.
Patents exist on the methods for preferential crystallization of the anhydrous form.144
There has been much confusion surrounding the occurrence of polymorphism for
phenylalanine. Multiple structure entries exist for enantiomerically pure phenylalanine
in the CSD.145 However, only one of these entries contains atomic coordinates65 (in this
case, determined for D-Phe, reference code SIMPEJ). The other entries (reference codes
QQQAUJ, QQQAUJ01 and QQQAUJ02), all reported by the same author,146,147 only
provide sets of unit cell parameters (unrelated to those of SIMPEJ), with QQQAUJ02
110
stated to be a redetermination of QQQAUJ. Significantly, the presence of multiple unit
cells seems to suggest the existence of polymorphism. It should be noted that recently
the crystal structure of enantiomerically pure Phe (reported Weissbuch et al.65) has been
called into question.148 King et al. have suggested that the previously reported
structure65 is of too high a symmetry, and propose that the symmetry is described by the
primitive space group P2 rather than the c-centred space group C2, based on the fact
that the diffraction pattern contains weak peaks that are unaccounted for by the C2
structure.
Work by Frey et al.,67 using solid-state 13C NMR, found that two samples of
L-Phe, one crystallized from water and the other crystallized from a water/ethanol
mixed solvent, had different spectra, indicating that they were different crystalline
forms. No identification of these phases was made, other than the report of their 13C
NMR spectra and the fact that they contain L-Phe.
The work described here was carried out with the aim of investigating the
existence of hydrate phases and the possibility of polymorphism in L-Phe. Any new
phases would subsequently be characterized and the crystal structures of these new
phases determined directly from powder XRD data. The observation of two distinct
crystalline forms by Frey et al. was used as the starting point for experimentation.
6.2 - Experimental Details
All experiments were carried out using a sample of L-Phe purchased from Sigma-
Aldrich (purity ≥ 98%), without further purification. This material was shown by
powder XRD to be a monophasic sample of the known (anhydrous) form of L-Phe.
Samples of L-Phe hydrates were crystallized from aqueous solution. Excess L-Phe
Figure 6.2: Molecular structure of aspartame, an artificial sweetener.
111
(above the solubility of L-Phe in water at 20 °C, ca. 50 mg of L-Phe per g of water) was
mixed with water and heated, using an oil bath, to 80 °C to allow for complete
dissolution, before removing from the bath and allowing to cool to ambient temperature
(at which the solution is supersaturated).
Powder XRD measurements were carried out on two Bruker D8 diffractometers:
one in transmission mode and one in reflection mode (both using Ge monochromated
Cu K1 radiation). For the purpose of “fingerprinting” the solid phases prepared in this
work, powder XRD measurements were performed in transmission mode and a
hygrometer was used to measure relative humidity (RH) levels within the diffractometer
cabinet. The temperature was held constant at 21 °C and the RH was found to vary
between 30% and 80% due to changes in atmospheric conditions. These measurements
were carried out on samples held flat between two pieces of tape (foil-type sample
holder). With this type of sample holder, the sample is not entirely sealed and may be
susceptible to changes in atmospheric conditions. High quality powder XRD data,
suitable for structure determination, were recorded on samples held within glass
capillaries. These capillaries were flame sealed and hence were isolated from the
external atmosphere, allowing the relative humidity level within the capillary to remain
constant. Powder XRD data were also recorded on a sample in reflection mode. In this
case, the sample was held within an indentation in a copper disk, seated in a silver-
coated copper cup. The sample was held at 294 K under vacuum, using an Oxford
Cryosystems Phenix temperature controller. This set-up allowed powder XRD data for
the sample to be measured under entirely anhydrous conditions.
Solid-state NMR measurements were carried out on an 850 MHz Bruker
AVANCE III spectrometer at the UK 850 MHz Solid-State NMR Facility (13C Larmor
frequency, 213.81 MHz, 12 kHz magic angle spinning frequency). High resolution 13C
solid-state NMR spectra were recorded using ramped 1H → 13C cross-polarization. DVS
measurements were carried out at 26.5 °C using a DVS Advantage 1 instrument
(Surface Measurement Systems Ltd., London, U.K.). The sample was placed in a glass
pan and held at a RH of 0% for two hours to allow an equilibrium moisture content to
be established. The RH was then changed incrementally in 5% steps every hour from 0
to 90% and then back to 0%. The mass of the sample was recorded as a function of time
and RH. TGA measurements were carried out using a Perkin-Elmer Pyris 6 instrument,
operated with a scan rate of 10 °C/min and a nitrogen purge gas.
112
6.3 - Results and Discussion
Visual examination of polycrystalline samples obtained by crystallization of L-Phe
from water revealed two distinctly different types of crystal morphology, suggesting the
possible existence of two different phases. In one case, the crystals were large and
block-like, while, in the other case, the entire solution appeared to crystallize as a single
solid mass composed of very fine fibre-like crystallites and containing a significant
amount of water. Formation of the fibrous morphology was found to be favoured by
crash cooling of the aqueous solution. Slow, controlled cooling using an incubator (e.g.,
from 85 °C to 25 °C, over twelve hours) was found to produce either the large block-
like crystals or the fine fibre-like crystallites (varying between batches) and in some
cases, both morphologies concomitantly.
Initially, powder XRD patterns were recorded for samples containing only
crystallites with one type of morphology (i.e., the samples either comprised entirely
block-like crystals or entirely fibrous crystals). It was found that the two morphologies
gave rise to distinct powder XRD patterns. On this basis, the block-like crystals were
identified as the known (anhydrous) phase of L-Phe. This phase was found to be
indefinitely stable under dry, ambient conditions. In contrast, for samples containing the
fibrous morphology, two different powder XRD patterns were observed. When the
powder XRD data were recorded using a sample held between two pieces of tape (and,
therefore, not entirely sealed from the atmosphere), the pattern observed at a given time
was found to correlate with the RH within the diffractometer cabinet and
interconversion between the two powder XRD patterns was observed occasionally. A
changeover between these two powder XRD patterns was observed at ca. 60% RH (at
21 °C). The observation of this change in powder XRD pattern with changing RH
indicates a change in solid phase and these phases are termed at this stage as a “low-
humidity” form and a “high-humidity” form. No change in the powder XRD pattern of
the block-like crystals (i.e., the known form of L-Phe) was observed with changes in
RH.
Both the low-humidity and high-humidity forms were found to be stable when
sealed in glass capillaries and high-quality powder XRD data were recorded for samples
held in this manner (Figure 6.3). The high-humidity form was maintained under high
humidity conditions prior to being packed into three glass capillaries with a small
113
amount of excess water (to ensure a high RH within the capillaries), before flame
sealing. The low-humidity form was kept in a desiccator containing silica gel drying
agent, before being packed into three glass capillaries and flame sealed. In each case,
the powder XRD patterns of the low-humidity form and high-humidity form recorded in
this manner matched those recorded below 60% RH and above 60% RH, respectively,
using the foil-type sample holder.
High-resolution solid-state 13C NMR spectra were recorded for the low-humidity
and high-humidity forms, and for the previously known anhydrous form of L-Phe. The
high-humidity form was stored in the presence of excess water until packed into the
NMR rotor, while the low-humidity form was stored with silica gel drying agent for
several days before packing into the NMR rotor. The NMR rotor was sealed during the
measurement on each sample and no evidence for interconversion between the high-
humidity and low-humidity forms was observed during acquisition of data. The solid-
state 13C NMR spectra are shown in Figure 6.4. From comparison of these spectra with
those reported by Frey et al.,67 it is clear that the sample crystallized from water was the
known anhydrous form of L-Phe, while the sample crystallized from water/ethanol
corresponds to the low-humidity form described in the present work.
The solid-state 13C NMR spectrum acquired for the known form of L-Phe in the
present work displays at least three distinct isotropic peaks for the -carbon
environment of the L-Phe molecule (Figure 6.5), suggesting that the crystal structure
contains at least three independent molecules of L-Phe. This result contrasts with the
crystal structure previously published for this form, which contains two independent
molecules of L-Phe. However, as mentioned in Section 6.1, King et al.148 have proposed
a new structure of lower symmetry for L-Phe. This structure contains four independent
molecules of L-Phe and is, thus, fully consistent with the 13C NMR spectrum recorded in
this work, though it cannot be definitively stated that other aspects of the structure
proposed by King et al.148 are confirmed by the NMR data (and, consequently, it cannot
be stated that this new structure is correct in its entirety). The solid-state 13C NMR data
for both the low-humidity and high-humidity forms indicates the presence of at least
two independent molecules of L-Phe within each of the crystal structures.
114
Figure 6.3: Powder XRD patterns recorded for (a) the previously known anhydrous
form (polymorph I) or L-Phe, (b) the high-humidity form (monohydrate) of L-Phe, (c)
the low-humidity form (hemihydrate) of L-Phe and (d) the new anhydrous form
(polymorph II) of L-Phe.
115
Figure 6.4: High resolution solid-state 13C NMR spectra of (a) the previously known
anhydrous form (polymorph I) of L-Phe, (b) L-Phe hemihydrate and (c) L-Phe
monohydrate. Spinning side bands are marked in red.
116
Figure 6.5: High resolution solid-state 13C NMR spectrum of the previously known
anhydrous form (polymorph I) of L-Phe, expanded to show the presence of at least three
peaks in the -carbon region of the spectrum.
To determine the water content of the high-humidity and low-humidity forms,
dynamic vapour sorption (DVS) measurements were carried out. The sample consisting
of crystallites of fibrous morphology (crystallized from water) was initially dried, under
vacuum, in the presence of phosphorus pentoxide. The results of the DVS experiment
are plotted in Figures 6.6 and 6.7. As the RH is increased from 0%, a rapid absorption
of water is observed. At ca. 5% RH, an uptake corresponding to ca. 0.25 equivalents of
water has occurred. With further increase in RH, there is a more gradual uptake of water
and by ca. 60%, close to 0.5 equivalents of water have been absorbed. Increasing the
RH further to 65%, a second stage of water absorption occurs, with an abrupt increase
in water uptake corresponding to a further 0.5 equivalents of water, i.e., a total increase
of 1 equivalent of water, compared to the material at 0% RH. On further increasing the
RH from 65% to 90%, there is essentially no further uptake of water.
On decreasing the RH from 90%, the events described previously occur in
reverse. It should be noted, however, that hysteresis is observed in the abrupt event that
corresponds to a decrease in water content from 1 equivalent to 0.5 equivalents. In the
117
desorption cycle, this event occurs at ca. 55% RH (compared to ca. 65% RH in the
absorption cycle).
By defining the stoichiometry of the observed phases as L-Phe(H2O)x, it is clear
from the DVS results that by starting from an anhydrous phase (x = 0) at 0% RH, a non-
stoichiometric hydrate with variable water content ranging from a quarter-hydrate to a
hemihydrate (0.25 ≤ x ≤ 0.5) exists at low RH (lower than 55 to 65% RH), while a
stoichiometric monohydrate (x = 1) exists at high RH. TGA analysis of the low-
humidity form (Figure 6.8) indicates the loss of ca. 0.5 equivalents of water upon
heating, confirming that the non-stoichiometric, low-humidity form includes a
hemihydrate as one end-member of the composition range. Henceforth, the low-
humidity form is referred to as the hemihydrate (non-stoichiometric, with actual water
content 0.25 ≤ x ≤ 0.5) and the high-humidity form is referred to as the monohydrate.
As noted above, no change in the powder XRD pattern of the previously known form of
anhydrous L-Phe is observed with changes in humidity. Thus, the phase present at 0%
RH in the DVS measurements must represent a new polymorph of L-Phe, henceforth
Figure 6.6: Dynamic vapour sorption (DVS) experiment demonstrating the hydration
behaviour of L-Phe. Shown in black is the stepwise change in relative humidity over
time. Shown in colour is the mass of the sample recorded on increasing (red) and
decreasing (blue) the relative humidity.
118
designated polymorph II. The previously known anhydrous form of L-Phe is designated
polymorph I.
The crystal structures of the hemihydrate and monohydrate were determined
directly from the powder XRD data recorded on the samples sealed within glass
capillaries. Powder XRD data on polymorph II of L-Phe were recorded by holding a
sample of the monohydrate under vacuum, within the Phenix temperature controller on
the powder X-ray diffractometer in reflection mode. By subjecting the sample to
vacuum (i.e., a very low humidity environment), polymorph II of L-Phe was formed.
Details on the structure determination process for the three new phases are given in
Section 6.4.
The three new crystal structures (polymorph II of L-Phe, L-Phe hemihydrate and
L-Phe monohydrate) share several notable features (Figure 6.9, Table 6.1). All three
Figure 6.7: A replotting of Figure 6.6, showing the change in sample mass with relative
humidity in the DVS experiment. The data recorded on increasing humidity are shown
in red and the data recorded on decreasing humidity are shown in blue. Represented by
dashed lines are the mass values for structures containing 0, 0.25, 0.5 and 1 equivalent
of water, i.e., anhydrous, quarter-hydrate, hemihydrate and monohydrate forms (taking
the lowest mass value during the experiment, 6.319 mg, to represent the anhydrous
form).
119
structures can be described in terms hydrogen-bonded bilayers parallel to the ab-plane,
containing the carboxylate and ammonium groups of the L-Phe molecules. In addition,
in the hemihydrate and monohydrate structures, the water molecules are located within
these hydrogen-bonded bilayers. The phenyl rings of the L-Phe molecules are located in
the region between adjacent hydrogen-bonded bilayers. In addition, all three structures
contain two crystallographically distinct molecules of L-Phe (as corroborated by solid-
state 13C NMR), with slightly different molecular conformations (the torsional angles in
the two distinct L-Phe molecules in each structure differ by up to ca. 20°).
Space
Group
a / Å b / Å c / Å ° V / Å3
Monohydrate P21 13.0075(4) 5.43017(11) 13.93996(33) 101.1125(14) 966.16(6)
Hemihydrate P21 12.1112(5) 5.42130(17) 13.7691(5) 100.0115(35) 890.29(8)
Polymorph II P21 12.063(11) 5.412(5) 13.676(13) 99.5976(26) 880.3(24)
Polymorph I C2 8.804 6.041 31.564 96.6 1667.6
Table 6.1: Unit cell parameters for L-Phe monohydrate, hemihydrate, anhydrous
polymorph II and the previously known anhydrous polymorph I.65
L-Phe hemihydrate and monohydrate can be described as channel hydrate
structures. In each of these structures, the water molecules are present within channels
parallel to the b-axis, located within the hydrogen-bonded bilayers. In the hemihydrate,
Figure 6.8: Thermogravimetric analysis (TGA) of the low-humidity form. Dashed lines
indicate the percentage mass difference expected for the loss of half an equivalent of
water from the structure.
120
there is only a single crystallographically distinct water molecule. The distance between
adjacent water molecules along the channel is too large to give rise to O–H···O
hydrogen bonding between them (O···O distance: 5.42 Å, corresponding to the unit cell
translation along the b-axis). On the other hand, the O···O distances (2.84 and 2.87 Å)
between the water molecule and oxygen atoms of carboxylate groups in two different
Figure 6.9: Crystal structures of (a) the monohydrate phase of L-Phe, (b) the
hemihydrate phase of L-Phe, and (c) the new anhydrous phase, polymorph II of L-Phe.
In each case the view along the b-axis is shown on the left and the view along the c-axis
is shown on the right (in this view, only one layer within the hydrogen-bonded bilayer is
shown, with the phenyl rings of the L-Phe molecules omitted for clarity). Only one
orientation of the hydrogen bonded water chain is shown in (a). In (a) and (b), the
water channel (running parallel to the b-axis) is highlighted in cyan, while in (c), the
empty channel is outlined in cyan. The helical hydrogen-bonded motif discussed in the
text is marked with an asterisk.
121
L-Phe molecules may be interpreted as O–H···O hydrogen bonds. However, the
OCOO···Owater···OCOO angle (162.6°) is too large to allow relatively linear O–H···O
hydrogen bonds to be formed simultaneously with both L-Phe molecules acting as
acceptors. It is therefore likely that at the local level (in both spatial and temporal
terms), each water molecule forms a strong and linear hydrogen bond with only one of
the two acceptors, with the other hydrogen atom on the water molecule not engaged in
any significant hydrogen-bonding interaction. There is likely to be significant disorder
in this arrangement and, due to the uncertainty in the time-averaged locations of the
hydrogen atoms of the water molecule, these hydrogen atoms were omitted from the
final Rietveld refinement. The lack of multiple, simultaneously formed, strong hydrogen
bonds involving water in this structure may explain the non-stoichiometry observed for
the hemihydrate phase, i.e., the water molecules can escape relatively easily from the
channel under low humidity conditions.
The monohydrate, as previously mentioned, also contains water molecules in
channels parallel to the b-axis. In this case, there are two distinct water molecules in the
crystal structure and they lie along the channels. The O···O distances between adjacent
water molecules along the channel are 2.82 and 2.98 Å, corresponding to
O–H···O hydrogen bonding between the water molecules. Thus, the water molecules
form an O–H···O–H···O–H chain along the b-axis. These chains may run in either
direction along the b-axis (i.e., either O–H···O–H···O–H or H–O···H–O···H–O), with
the two possibilities only differing in the hydrogen atom positions (the positions of the
oxygen atoms are identical in each case). The final Rietveld refinement of the
monohydrate invoked disorder of these two orientations of the hydrogen bonded chain,
with fractional occupancies of xH and 1 – xH for the hydrogen atoms in the positive and
negative directions of the chain, respectively. During refinement, these occupancies did
not depart significantly from the initial value of 0.5, indicating that the populations of
each chain orientation are roughly equal. Each water molecule is also engaged in
hydrogen bonding to L-Phe molecules. One water molecule is able to act as both the
donor in an O–H···O hydrogen bond (O···O distance: 2.61 Å) to the carboxylate group
of one L-Phe molecule and as an acceptor within an O···H–N hydrogen bond (O···N
distance: 2.79 Å) to the ammonium group of a second L-Phe molecule. The other water
molecule, however, only forms a single hydrogen bond to an L-Phe molecule, acting as
the donor within an O–H···O bond (O···O distance: 2.81 Å) to an oxygen atom in a
carboxylate group.
122
Another notable difference between the structures of the monohydrate and
hemihydrate concerns the nature of the helical hydrogen-bonded motif that follows the
21 screw axis (parallel to the b-axis) in each structure. In the hemihydrate, this helical
motif involves solely N–H···O hydrogen bonds between L-Phe molecules in the two
layers of the hydrogen-bonded bilayer while, in the monohydrate, the L-Phe molecules
are instead involved in N–H···O–H···OCOOH hydrogen bonds, where the O–H bond of
an intervening water molecule forms part of the helical motif.
The crystal structure of polymorph II of L-Phe is very similar to that of the
hemihydrate and there are only minor displacements in the positions of the L-Phe
molecules between the two structures. Due to the fact that differences between the two
structures are only minor, polymorph II of L-Phe may be referred to as an “isomorphic
desolvate”149 of the hemihydrate. The presence of channel structures occupied by water
in the monohydrate and hemihydrate, and the fact that the empty channels remain
without collapse in polymorph II of anhydrous L-Phe, are completely consistent with the
facile water absorption/desorption observed within this set of materials. It should be
noted that the density of polymorph II of L-Phe (1.25 g cm–1) is significantly lower than
that of polymorph I of L-Phe (1.32 g cm–1), which may suggest that polymorph II is less
thermodynamically stable than polymorph I. The lower density of polymorph II is
related to the presence of the empty channel structure, which is not present in the
structure of polymorph I. The conversion of polymorph II to polymorph I has not been
observed during this work (on the timescale of several days), indicating a kinetic barrier
to the transformation. Exposure of polymorph II to the atmosphere leads to rapid and
facile uptake of water (even at low RH, as shown in the DVS experiments), to give the
hemihydrate or monohydrate depending on the humidity level.
Channel hydrates are a well-known class of hydrate structure and examples of
such materials have been discussed in the literature.150-153 Particularly relevant is the
hydrate crystal structure of another amino acid, DL-proline monohydrate.154 In this
crystal structure, water molecules are located within channels and are hydrogen bonded
both to each other and to the carboxylate groups of neighbouring proline molecules,
strongly resembling the features of the structure of L-Phe monohydrate.
123
6.4 - Details of Structure Determination from Powder XRD Data
6.4.1 - Structure Determination of L-Phe Hemihydrate
The powder XRD pattern of L-Phe hemihydrate was indexed using the program
TREOR113, within the CRYSFIRE106 program suite, with monoclinic metric symmetry
and dimensions (following transformation into the standard form with the program
PLATON155,156): a = 12.13 Å, b = 5.42 Å, c = 13.78 Å, β = 100.1° (V = 892.3 Å3). From
systematic absences, the space group was assigned as P21 and from density arguments
the asymmetric unit was determined to contain two molecules of L-Phe (and therefore
one of water). Le Bail fitting was performed on the powder XRD pattern, using the
program GSAS,107 achieving a high quality of fit (Figure 6.10(a), Rwp = 2.67%,
Rp = 2.01%) and the refined cell and profile parameters were used in the subsequent
structure solution calculations.
The program EAGER was used to carry out structure solution with 16
independent GA calculations performed. In total, 200 generations were carried out with
a population of 200 structures and a total of 20 mating operations and 100 mutation
operations were carried out per generation. One L-Phe molecule was represented by
eight structural variables: two positional variables (in the space group P21, the origin
can be fixed arbitrarily along the b-axis), three orientational variables and three
torsional variables (Figure 6.11); the second L-Phe molecule was defined similarly but
with three positional variables, allowing movement along the b-axis. The water
molecule was defined using three positional and three orientational variables. Both
L-Phe molecules were defined as zwitterions during structure solution. After 200
generations, the same good quality fit was found in eight of the 16 calculations and this
was used as the starting point for Rietveld refinement.
Rietveld refinement was carried out using the program GSAS.107 Standard
restraints were applied to bond lengths and angles and planar restraints were used to
maintain the planarity of the aromatic ring and the carboxylate group. These restraints
were relaxed during the refinement procedure. The hydrogen-bonding network between
ammonium and carboxylate groups was loosely restrained based on optimum hydrogen-
bonding geometries. Intermolecular bond-length restraints were used to orient the
ammonium group in order to give the optimum hydrogen-bonding environment. Three
separate Uiso values were refined (one for each molecule) with the value of Uiso for the
124
Figure 6.10: (a) Le Bail fit and (b) Rietveld fit for powder XRD data recorded on L-Phe
hemihydrate. The experimental data is shown as red crosses, the calculated fit as a
green line and the difference between the two as a magenta line. Peak positions are
displayed as black tick marks.
125
Figure 6.11: The three torsional angles of L-Phe that were allowed to vary in the
structure solution calculations.
hydrogen atoms set to be 1.2 times that of the non-hydrogen atoms. Due to uncertainty
in the time-averaged positions of the hydrogen atoms of the water molecules in this
structure (caused by the likelihood of significant disorder), the hydrogen atoms of the
water molecule were omitted from the final Rietveld refinement. Due to the non-
stoichiometry of the hemihydrate, the occupancy of the water molecule was allowed to
vary. However, the occupancy was found to refine to a value very close to 1 and so was
fixed at 1 in the final refinement. No correction for preferred orientation was used. The
final Rietveld refinement gave a high quality fit to the powder XRD data [Figure
6.10(b), Rwp = 2.96%, Rp = 2.21%] with the following refined parameters:
a = 12.1112(5) Å, b = 5.42130(17) Å, c = 13.7691(5) Å, β = 100.0115(35)°;
V = 890.29(8) Å3 (space group P21; 2θ range, 4° to 70°; 3867 profile points; 170 refined
variables).
6.4.2 - Structure Determination of L-Phe Monohydrate
The powder XRD pattern of L-Phe monohydrate was indexed using the program
DICVOL91157, within the CRYSFIRE106 program suite, as a monoclinic cell and
dimensions: a = 13.05 Å, b = 5.44 Å, c = 13.98 Å, β = 101.2° (V = 973.5 Å3). From
systematic absences, the space group was assigned as P21 and from density arguments
the asymmetric unit was determined to contain two molecules of L-Phe (and therefore
two of water). Le Bail fitting was performed on the powder XRD pattern, using the
program GSAS,107 achieving a high quality of fit [Figure 6.12(a), Rwp = 1.41%,
Rp = 1.06%] and the refined cell and profile parameters were used in the subsequent
structure solution calculations.
The program EAGER was used to carry out structure solution with 16
independent GA calculations performed. In total, 400 generations were carried out with
a population of 200 structures and a total of 20 mating operations and 100 mutation
126
operations were carried out per generation. One L-Phe molecule was represented by
eight structural variables: two positional variables (in the space group P21, the origin
can be fixed arbitrarily along the b-axis), three orientational variables and three
torsional variables (Figure 6.11); the second L-Phe molecule was defined similarly but
with three positional variables, allowing movement along the b-axis. Each water
molecule was defined using three positional and three orientational variables. Both
L-Phe molecules were defined as zwitterions during structure solution. After 400
generations, the same good quality fit was found in four of the 16 calculations and this
structure was used as the starting point for Rietveld refinement.
Rietveld refinement was carried out using the program GSAS.107 Standard
restraints were applied to bond lengths and angles and planar restraints were used to
maintain the planarity of the aromatic ring and the carboxylate group. These restraints
were relaxed during the refinement procedure. The hydrogen-bonding network between
ammonium and carboxylate groups was loosely restrained based on optimum hydrogen-
bonding geometries. Intermolecular bond-length restraints were used to orient the
ammonium group in order to give the optimum hydrogen-bonding environment. Four
separate Uiso values were refined (one for each molecule) with the value of Uiso for the
hydrogen atoms set to be 1.2 times that of the non-hydrogen atoms. Disorder of the two
possible orientations of the O–H···O–H···O–H hydrogen-bonded chain along the b-axis
was included with fractional occupancies of xH and 1 – xH for the hydrogen atoms in the
positive and negative of directions of the chain, respectively. During refinement, these
occupancies were not found to vary significantly from 0.5 and so each occupancy was
fixed to 0.5 in the final refinement. No correction for preferred orientation was used.
The final Rietveld refinement gave a high quality fit to the powder XRD data [Figure
6.12(b), Rwp = 1.60%, Rp = 1.18%], with the following refined parameters:
a = 13.0075(4) Å, b = 5.43017(11) Å, c = 13.93996(33) Å, β = 101.1125(14)°;
V = 966.16(6) Å3 (space group P21; 2θ range, 4° to 70°; 3867 profile points; 183 refined
variables).
127
Figure 6.12: (a) Le Bail fit and (b) Rietveld fit for powder XRD data recorded on L-Phe
monohydrate. The experimental data is shown as red crosses, the calculated fit as a
green line and the difference between the two as a magenta line. Peak positions are
displayed as black tick marks.
128
6.4.3 - Structure Determination of Polymorph II of L-Phe
The structure of polymorph II of L-Phe of was determined using the structure of
the hemihydrate as a starting point. Le Bail fitting to the powder XRD data gave a good
quality of fit [Figure 6.13(a), Rwp = 3.38%, Rp = 2.35%]. During Rietveld refinement,
the occupancy of the water molecule from the hemihydrate phase was allowed to vary.
This occupancy was found to refine to a small, negative value and so was subsequently
fixed at zero, corresponding to an anhydrous phase. Preferred orientation was taken into
account using the March-Dollase function114,115 and a correction for surface roughness
was included using a normalized form of the function proposed by Suortti.158 Two
separate Uiso values were refined (one for each molecule) with the value of Uiso for the
hydrogen atoms set to be 1.2 times that of the non-hydrogen atoms. The final Rietveld
refinement gave a good fit to the data [Figure 6.13(b), Rwp = 4.50%, Rp = 3.01%] with
the following refined parameters: a = 12.063(11) Å, b = 5.412(5) Å, c = 13.676(13) Å,
β = 99.5976(26)°; V = 880.3(24) Å3 (space group P21; 2θ range, 6.25° to 41.75°; 2196
profile points; 160 refined variables).
6.5 - Conclusions
The characterization of three new solid phases of L-Phe, including two hydrate
forms and a new polymorph of anhydrous L-Phe, along with the rationalization of the
transformation processes between them, add clarity to the previously confused topic of
the solid-state forms of L-Phe. The monohydrate and hemihydrate structures determined
in this work represent the first reported crystal structures of hydrated forms of L-Phe.
Analysis of the CSD indicates that, prior to the present work, five of the 20 directly
genetically encoded amino acids (arginine,159 asparagine,160 aspartic acid,161 proline162
and serine163) were known to form hydrate structures as the enantiomerically pure
amino acid. Four cases (serine,164 arginine,165,166 glutamic acid167 and proline154,168) of
hydrate structures for the racemic amino acids have been reported (with both a
monohydrate166 and a dihydrate165 reported for DL-arginine). L-Phe represents the first
enantiomerically pure amino acid for which more than one hydrate phase has been
discovered and structurally characterized. In addition, the discovery of polymorphism
within L-Phe represents the sixth case within the 20 directly genetically encoded amino
acids for which polymorphism has been confirmed by crystal structure determination.
129
Figure 6.13: (a) Le Bail fit and (b) Rietveld fit for powder XRD data recorded on
Polymorph II of L-Phe. The experimental data is shown as red crosses, the calculated fit
as a green line and the difference between the two as a magenta line. Peak positions are
displayed as black tick marks.
130
Chapter 7: Future Work
While the work in this thesis progresses the current knowledge on several specific
organic solid-state systems, many questions are raised which must, in the future, be
answered. These issues and the possibilities for future work are now discussed.
In the work conducted on m-ABA in Chapter 3, three new polymorphs were
discovered and characterized (including crystal structure determination of each of these
new polymorphs directly from powder XRD data). However, the crystal structure of the
previously reported Form I of m-ABA remains unknown. Future work must focus on
determination of the crystal structure of Form I, with special attention given to the
currently intractable indexing stage. Possible avenues of this study include recording
further powder XRD data at different temperatures (with the aim of reducing the high
degree of peak overlap found at ambient temperature in the region 20° ≤ 2 ≤ 35°) and,
perhaps, the use of other experimental techniques, e.g., electron diffraction. In addition,
the relative thermodynamic stabilities of Form I and Form IV are currently unknown, as
slurry experiments result in the formation of Form III, the most stable (currently known)
polymorph of m-ABA at ambient temperature. The CLASSIC NMR technique
developed during the work presented in Chapter 3 represents a powerful new tool in the
study of crystallization processes. By allowing the evolution of both the solid and liquid
phases to be studied during a single crystallization experiment, changes in both the solid
forms present (including the potential to discover novel, transient forms) and in the
aggregation behaviour of the solution may be observed. It may be expected that the
CLASSIC NMR technique will, in the future, be applied to reveal significant new
insights on a range of systems of interest.
In Chapter 4, up to three new solid phases of TPPO were discovered using the
CLASSIC NMR method. While a new solid phase of TPPO was subsequently isolated
during crystallization experiments conducted within sealed glass capillaries and
identified as a new phase by powder XRD, the crystal structure of this phase remains
undetermined. Indexing of the powder XRD of this new phase appeared to be
successful; however, structure solution using EAGER has not yet progressed to finding
a suitable candidate structure for Rietveld refinement. Further work must be conducted
to determine the crystal structure of this new phase. A remaining unknown for TPPO is
the methods for crystallization of the monoclinic II and monoclinic III polymorphs. To
131
conclusively rule out the possibility that one or two of the unknown phases observed in
the CLASSIC NMR study of the crystallization of TPPO may be the monoclinic II or
monoclinic III polymorph, these polymorphs must be crystallized and solid-state 31P
NMR spectra acquired. Two previously unknown transient phases of MDPPO were
discovered during the crystallization of MDPPO from toluene through use of in-situ
solid-state 31P NMR. The identity of these new phases is currently unknown and so
work must focus on the isolation and characterization of these phases. During the
crystallization, only small amounts of these two new phases were formed concomitantly
with a large amount of the only previously known phase of MDPPO. New
crystallization conditions will likely need to be discovered in order to isolate the two
new phases.
During the work described in Chapter 5, it was discovered that, contrary to the
prior literature,22 crystallization of Form II of rac-ibuprofen does not require cooling of
supercooled liquid rac-ibuprofen to a temperature lower than the glass transition,
followed by annealing at 258 K. This work has demonstrated that directly cooling liquid
rac-ibuprofen to 258K (or even 273 K) and maintaining the sample at that temperature
for an extended length of time is sufficient for crystallization of Form II. During the
DSC experiments, issues with reproducibility can be seen between experiments, i.e., the
same temperature regime may produce very different results. This irreproducibility may
be caused by a variable lag time before crystallization of Form II of rac-ibuprofen,
which, as shown in the powder XRD experiments, may be considerably longer than the
timescale of the DSC experiments. Further work should be conducted to investigate and
explain the problems with reproducibility.
As described in Chapter 6, a new polymorph (polymorph II) of L-Phe and two
hydrate phases (a monohydrate and a non-stoichiometric hydrate with quarter-hydrate
and hemihydrate end members) were discovered and structurally characterized in the
present work. All three crystal structures share similarities, with water molecules
present within channels in the structures of the two hydrate phases. Polymorph II of
L-Phe also contains these channels, empty of water molecules in this case, and thus
polymorph II represents an “isomorphic desolvate”. The existence of these channels
appears to be the reason for the facile and completely reversible changes between the
monohydrate, hemihydrate and (anhydrous) polymorph II. It would be interesting to
attempt the preparation of a single crystal of L-Phe monohydrate to investigate whether
132
the transformation between the monohydrate, hemihydrate and polymorph II of L-Phe
occurs in a single crystal to single crystal manner, without loss of crystal integrity. A
question also remains over the thermodynamic stability of polymorph II with respect to
the previously known polymorph I of L-Phe. The density of polymorph II (1.25 g cm–1)
is significantly lower than that of polymorph I (1.32 g cm–1) which, according to the
density rule, may suggest a lower thermodynamic stability. However, no transformation
between the two polymorphs was observed over the course of the experiments (a
timescale of several days) conducted in this work.
133
References
[1] L. Yu, G. A. Stephenson, C. A. Mitchell, C. A. Bunnell, S. V. Snorek, J. J.
Bowyer, T. B. Borchardt, J. G. Stowell and S. R. Byrn, J. Am. Chem. Soc., 122,
585-591 (2000).
[2] S. Chen, I. A. Guzei and L. Yu, J. Am. Chem. Soc., 127, 9881-9885 (2005).
[3] L. A. Yu, Acc. Chem. Res., 43, 1257-1266 (2010).
[4] G. L. Perlovich, L. K. Hansen and A. Bauer-Brandl, J. Therm. Anal. Calorim.,
66, 699-715 (2001).
[5] E. V. Boldyreva, V. A. Drebushchak, T. N. Drebushchak, I. E. Paukov, Y. A.
Kovalevskaya and E. S. Shutova, J. Therm. Anal. Calorim., 73, 409-418 (2003).
[6] E. V. Boldyreva, V. A. Drebushchak, T. N. Drebushchak, I. E. Paukov, Y. A.
Kovalevskaya and E. S. Shutova, J. Therm. Anal. Calorim., 73, 419-428 (2003).
[7] J. Bauer, S. Spanton, R. Henry, J. Quick, W. Dziki, W. Porter and J. Morris,
Pharm. Res., 18, 859-866 (2001).
[8] D. W. Schacht, M. Hannay and H.-M. Wolff, Patent No. US8246979 (B2);
US2005079206 (A1) (14th April 2005).
[9] H.-M. Wolff, L. Quere and J. Riedner, Patent No. US2010311806 (A1) (9th
December 2010).
[10] C. R. Groom, CPOSS, University College London (2011).
[11] GLAXO INC. v. NOVOPHARM LTD., 830 F.Supp. 869 (1993), United States
District Court, E.D. North Carolina, Raleigh Division
[12] GLAXO INC. v. NOVOPHARM LTD., 52 F.3d 1043 (1995), United States
Court of Appeals, Federal Circuit.
[13] M. Haïsa, S. Kashino and H. Maeda, Acta Crystallogr. Sect. B, 30, 2510-2512
(1974).
[14] M. Haïsa, S. Kashino, R. Kawai and H. Maeda, Acta Crystallogr. Sect. B, 32,
1283-1285 (1976).
[15] M.-A. Perrin, M. A. Neumann, H. Elmaleh and L. Zaske, Chem. Commun.,
3181-3183 (2009).
[16] P. J. Wheatley, J. Chem. Soc., S, 6036-6048 (1964).
[17] P. Vishweshwar, J. A. McMahon, M. Oliveira, M. L. Peterson and M. J.
Zaworotko, J. Am. Chem. Soc., 127, 16802-16803 (2005).
[18] A. D. Bond, R. Boese and G. R. Desiraju, Angew. Chem. Int. Ed., 46, 615-617
(2007).
[19] A. D. Bond, R. Boese and G. R. Desiraju, Angew. Chem. Int. Ed., 46, 618-622
(2007).
[20] A. D. Bond, K. A. Solanko, S. Parsons, S. Redder and R. Boese,
CrystEngComm, 13, 399-401 (2010).
134
[21] J. F. McConnell, Cryst. Struct. Commun., 3, 73-75 (1974).
[22] E. Dudognon, F. Danede, M. Descamps and N. T. Correia, Pharm. Res., 25,
2853-2858 (2008).
[23] V. López-Mejías, J. W. Kampf and A. J. Matzger, J. Am. Chem. Soc., 134, 9872-
9875 (2012).
[24] A. L. Grzesiak, M. D. Lang, K. Kim and A. J. Matzger, J. Pharm. Sci., 92, 2260-
2271 (2003).
[25] J. Anwar, S. E. Tarling and P. Barnes, J. Pharm. Sci., 78, 337-342 (1989).
[26] F. C. Chan, J. Anwar, R. Cernik, P. Barnes and R. M. Wilson, J. Appl.
Crystallogr., 32, 436-441 (1999).
[27] J. W. Mullin, Crystallization. (Butterworth & Co. Ltd., London, 1961).
[28] M. C. Etter, J. Phys. Chem., 95, 4601-4610 (1991).
[29] R. I. Walton and D. O'Hare, Chem. Commun., 2283-2291 (2000).
[30] R. J. Francis, S. O'Brien, A. M. Fogg, P. S. Halasyamani, D. O'Hare, T. Loiseau
and G. Férey, J. Am. Chem. Soc., 121, 1002-1015 (1999).
[31] I. R. Evans, J. A. K. Howard, T. Sreckovic and M. M. Ristic, Mater. Res. Bull.,
38, 1203-1213 (2003).
[32] M. Berglund and M. E. Wieser, Pure Appl. Chem., 83, 397-410 (2011).
[33] C. Riekel and R. Schöllhorn, Mater. Res. Bull., 11, 369-376 (1976).
[34] J. Schöll, D. Bonalumi, L. Vicum, M. Mazzotti and M. Muller, Cryst. Growth
Des., 6, 881-891 (2006).
[35] M. P. Feth, J. Jurascheck, M. Spitzenberg, J. Dillenz, G. Bertele and H. Stark, J.
Pharm. Sci., 100, 1080-1092 (2011).
[36] G. Févotte, Chem. Eng. Res. Des., 85, 906-920 (2007).
[37] P.-G. Jönsson and Å. Kvick, Acta Crystallogr. Sect. B, 28, 1827-1833 (1972).
[38] Y. Iitaka, Acta Crystallogr., 13, 35-45 (1960).
[39] Å. Kvick, W. M. Canning, T. F. Koetzle and G. J. B. Williams, Acta
Crystallogr. Sect. B, 36, 115-120 (1980).
[40] E. V. Boldyreva, S. N. Ivashevskaya, H. Sowa, H. Ahsbahs and H.-P. Weber, Z.
Kristallogr., 220, 50-57 (2005).
[41] A. Dawson, D. R. Allan, S. A. Belmonte, S. J. Clark, W. I. F. David, P. A.
McGregor, S. Parsons, C. R. Pulham and L. Sawyer, Cryst. Growth Des., 5,
1415-1427 (2005).
[42] C. E. Hughes and K. D. M. Harris, J. Phys. Chem. A, 112, 6808-6810 (2008).
[43] C. E. Hughes and K. D. M. Harris, Chem. Commun., 46, 4982-4984 (2010).
[44] J. M. Shi, M. W. Anderson and S. W. Carr, Chem. Mater., 8, 369-375 (1996).
[45] C. Allais, G. Keller, P. Lesieur, M. Ollivon and F. Artzner, J. Therm. Anal.
Calorim., 74, 723-728 (2003).
[46] C. J. Brown and M. Ehrenberg, Acta Crystallogr. Sect. C, 41, 441-443 (1985).
135
[47] C. D. G. Boone, J. L. Derissen and J. C. Schoone, Acta Crystallogr. Sect. B, 33,
3205-3206 (1977).
[48] H. Takazawa, S. Ohba and Y. Saito, Acta Crystallogr. Sect. C, 42, 1880-1881
(1986).
[49] M. Svärd, F. L. Nordstrom, T. Jasnobulka and A. C. Rasmuson, Cryst. Growth
Des., 10, 195-204 (2010).
[50] T. Yoshida and S. Sawada, Bull. Chem. Soc. Jpn., 49, 3319-3320 (1976).
[51] J. Voogd, B. H. M. Verzijl and A. J. M. Duisenberg, Acta Crystallogr. Sect. B,
36, 2805-2806 (1980).
[52] T. F. Lai and R. E. Marsh, Acta Crystallogr., 22, 885-893 (1967).
[53] M. Alleaume, Salascim.G and J. Decap, C. R. Seances Acad. Sci., Ser. C, 262,
416-417 (1966).
[54] R. C. G. Killean, P. Tollin, D. G. Watson and D. W. Young, Acta Crystallogr.,
19, 482-483 (1965).
[55] M. Kuhnert-Brandstaetter and H. Grimm, Mikrochim. Acta, 1208-1209 (1969).
[56] M. C. Etter and P. W. Baures, J. Am. Chem. Soc., 110, 639-640 (1988).
[57] C. P. Brock, W. B. Schweizer and J. D. Dunitz, J. Am. Chem. Soc., 107, 6964-
6970 (1985).
[58] G. Bandoli, Bortoloz.G, D. A. Clemente, U. Croatto and Panatton.C, J. Chem.
Soc. A, 2778-2780 (1970).
[59] A. L. Spek, Acta Crystallogr. Sect. C, 43, 1233-1235 (1987).
[60] F. Dornhaus, M. Bolte, H.-W. Lerner and M. Wagner, Eur. J. Inorg. Chem.,
5138-5147 (2006).
[61] D. H. Ryu and E. J. Corey, J. Am. Chem. Soc., 127, 5384-5387 (2005).
[62] Z. V. Belyakova, M. G. Pomerantseva, L. A. Efimova, E. A. Chernyshev and P.
A. Storozhenko, Russ. J. Gen. Chem., 80, 728-733 (2010).
[63] R. M. Denton, J. An, B. Adeniran, A. J. Blake, W. Lewis and A. M. Poulton, J.
Org. Chem., 76, 6749-6767 (2011).
[64] P. Derollez, E. Dudognon, F. Affouard, F. Danede, N. T. Correia and M.
Descamps, Acta Crystallogr. Sect. B, 66, 76-80 (2010).
[65] I. Weissbuch, F. Frolow, L. Addadi, M. Lahav and L. Leiserowitz, J. Am. Chem.
Soc., 112, 7718-7724 (1990).
[66] R. Mohan, K. K. Koo, C. Strege and A. S. Myerson, Ind. Eng. Chem. Res., 40,
6111-6117 (2001).
[67] M. H. Frey, J. A. DiVerdi and S. J. Opella, J. Am. Chem. Soc., 107, 7311-7315
(1985).
[68] W. L. Bragg, Proc. R. Soc. London, Ser. A, 89, 248-277 (1913).
[69] P. A. Doyle and P. S. Turner, Acta Crystallogr. Sect. A, 24, 390-397 (1968).
[70] M. F. C. Ladd and P. R. A. eds., Theory and Practice of Direct Methods in
Crystallography. (Plenum Press, New York, 1980).
136
[71] M. M. Woolfson, Acta Crystallogr. Sect. A, 43, 593-612 (1987).
[72] A. L. Patterson, Phys. Rev., 46, 0372-0376 (1934).
[73] G. Oszlányi and A. Sütő, Acta Crystallogr. Sect. A, 60, 134-141 (2004).
[74] G. Oszlányi and A. Sütő, Acta Crystallogr. Sect. A, 61, 147-152 (2005).
[75] C. Baerlocher, L. B. McCusker and L. Palatinus, Z. Kristallogr., 222, 47-53
(2007).
[76] K. D. M. Harris, M. Tremayne, P. Lightfoot and P. G. Bruce, J. Am. Chem. Soc.,
116, 3543-3547 (1994).
[77] A. Le Bail, H. Duroy and J. L. Fourquet, Mater. Res. Bull., 23, 447-452 (1988).
[78] G. S. Pawley, J. Appl. Crystallogr., 14, 357-361 (1981).
[79] P. Thompson, D. E. Cox and J. B. Hastings, J. Appl. Crystallogr., 20, 79-83
(1987).
[80] L. W. Finger, D. E. Cox and A. P. Jephcoat, J. Appl. Crystallogr., 27, 892-900
(1994).
[81] B. M. Kariuki, K. Psallidas, K. D. M. Harris, R. L. Johnston, R. W. Lancaster, S.
E. Staniforth and S. M. Cooper, Chem. Commun., 1677-1678 (1999).
[82] E. Tedesco, G. W. Turner, K. D. M. Harris, R. L. Johnston and B. M. Kariuki,
Angew. Chem. Int. Ed., 39, 4488-4491 (2000).
[83] D. Albesa-Jové, B. M. Kariuki, S. J. Kitchin, L. Grice, E. Y. Cheung and K. D.
M. Harris, ChemPhysChem, 5, 414-418 (2004).
[84] F. Guo and K. D. M. Harris, J. Am. Chem. Soc., 127, 7314-7315 (2005).
[85] Z. G. Pan, M. C. Xu, E. Y. Cheung, K. D. M. Harris, E. C. Constable and C. E.
Housecroft, J. Phys. Chem. B, 110, 11620-11623 (2006).
[86] F. Guo, J. Martí-Rujas, Z. Pan, C. E. Hughes and K. D. M. Harris, J. Phys.
Chem. C, 112, 19793-19796 (2008).
[87] H. M. Rietveld, J. Appl. Crystallogr., 2, 65-71 (1969).
[88] M. J. Duer ed., Solid-State NMR Spectroscopy Principles and Applications.
(Blackwell Science, 2002).
[89] A. Pines, M. G. Gibby and J. S. Waugh, J. Chem. Phys., 59, 569-590 (1973).
[90] B. M. Fung, A. K. Khitrin and K. Ermolaev, J. Magn. Reson., 142, 97-101
(2000).
[91] T. A. Carlson, Photoelectron and Auger Spectroscopy. (Plenum Press, London,
1975).
[92] G. Höhne, W. F. Hemminger and H.-J. Flammersheim, Differential Scanning
Calorimetry. (Springer, 2003).
[93] B. Shah, V. K. Kakumanu and A. K. Bansal, J. Pharm. Sci., 95, 1641-1665
(2006).
[94] A. D. Roman-Gutierrez, S. Guilbert and B. Cuq, J. Cereal Sci., 36, 347-355
(2002).
[95] K. Klokkers, United States Patent No. 6204255 (2001).
137
[96] C. J. Brown, Proc. R. Soc. London, Ser. A, 302, 185-199 (1968).
[97] A. Théorêt, Spectrochim. Acta, Part A, 27, 11-18 (1971).
[98] M. A. Palafox, M. Gil and J. L. Nunez, Vib. Spectrosc, 6, 95-105 (1993).
[99] G. Reinhard, M. Radtke and U. Rammelt, Corros. Sci., 33, 307-313 (1992).
[100] A. A. Al-Suhybani, Y. H. Sultan and W. A. Hamid, Materialwiss.
Werkstofftech., 22, 301-307 (1991).
[101] G. C. Levy, A. D. Godwin, J. M. Hewitt and C. Sutcliffe, J. Magn. Reson., 29,
553-562 (1978).
[102] L. Van Blaricom and E. F. Gilbert, J. Am. Chem. Soc., 61, 3238-3239 (1939).
[103] L. Gopal, C. I. Jose and A. B. Biswas, Spectrochim. Acta, Part A, 23, 513-518
(1967).
[104] I. Mentré, Ann. Chim. (Paris), 8, 115-122 (1973).
[105] L. M. Schwartz, R. I. Gelb, J. Mumfordzisk and D. A. Laufer, J. Chem. Soc.,
Perkin Trans. 2, 453-460 (1987).
[106] R. Shirley, The CRYSFIRE System for Automatic Powder Indexing: User’s
Manual. (The Lattice Press, Guildford, U.K., 1999).
[107] A. C. Larson and R. B. Von Dreele, Los Alamos National Laboratory Report,
LAUR 86-748 (2004).
[108] B. H. Toby, J. Appl. Crystallogr., 34, 210-213 (2001).
[109] G. Metz, X. L. Wu and S. O. Smith, J. Magn. Reson. A, 110, 219-227 (1994).
[110] M. C. Etter, Acc. Chem. Res., 23, 120-126 (1990).
[111] M. C. Etter, J. C. Macdonald and J. Bernstein, Acta Crystallogr. Sect. B, 46,
256-262 (1990).
[112] J. Bernstein, R. E. Davis, L. Shimoni and N. L. Chang, Angew. Chem. Int. Ed.,
34, 1555-1573 (1995).
[113] P. E. Werner, L. Eriksson and M. Westdahl, J. Appl. Crystallogr., 18, 367-370
(1985).
[114] A. March, Z. Kristallogr., 81, 285-297 (1932).
[115] W. A. Dollase, J. Appl. Crystallogr., 19, 267-272 (1986).
[116] J. W. Visser, J. Appl. Crystallogr., 2, 89-95 (1969).
[117] T. Charpentier, Solid State Nucl. Magn. Reson., 40, 1-20 (2011).
[118] R. K. Harris, R. E. Wasylishen and M. J. Duer, NMR Crystallography. (John
Wiley & Sons, Chichester, 2012).
[119] D. V. Dudenko, J. R. Yates, K. D. M. Harris and S. P. Brown, CrystEngComm,
15, 8797-8807 (2013).
[120] I. Bertini, G. Gallo, M. Korsak, C. Luchinat, J. Mao and E. Ravera,
ChemBioChem, 14, 1891-1897 (2013).
[121] D. Courtier-Murias, H. Farooq, H. Masoom, A. Botana, R. Soong, J. G.
Longstaffe, M. J. Simpson, W. E. Maas, M. Fey, B. Andrew, J. Struppe, H.
138
Hutchins, S. Krishnamurthy, R. Kumar, M. Monette, H. J. Stronks, A. Hume and
A. J. Simpson, J. Magn. Reson., 217, 61-76 (2012).
[122] S. E. Denmark, R. C. Smith and S. A. Tymonko, Tetrahedron, 63, 5730-5738
(2007).
[123] O. Mitsunobu and M. Yamada, Bull. Chem. Soc. Jpn., 40, 2380-2382 (1967).
[124] H. Staudinger and J. Meyer, Helv. Chim. Acta, 2, 635-646 (1919).
[125] G. Wittig and U. Schöllkopf, Chem. Ber. Recl., 87, 1318-1330 (1954).
[126] G. Wittig and W. Haag, Chem. Ber. Recl., 88, 1654-1666 (1955).
[127] Private communication submitted to CSD: A. T. H. Lenstra, Refcode:
TPEPHO12 (2007).
[128] P. W. Baures and J. V. Silverton, Acta Crystallogr. Sect. C, 46, 715-717 (1990).
[129] P. W. Baures, Acta Crystallogr. Sect. C, 47, 2715-2716 (1991).
[130] S. W. Ng, Acta Crystallogr. Sect. E, 65, o1431 (2009).
[131] Private communication submitted to CSD: J. Wahl, M. Bolte and D. Franz,
Refcode: OBEBAA (2011).
[132] S. Arumugam, C. Glidewell and K. D. M. Harris, J. Chem. Soc., Chem.
Commun., 724-726 (1992).
[133] F. Kohlbeck and E. M. Hörl, J. Appl. Crystallogr., 11, 60-61 (1978).
[134] WHO. Model Lists of Essential Medicines, 18th ed.; 2013.
[135] N. M. Davies, Clin. Pharmacokinet., 34, 101-154 (1998).
[136] K. D. Rainsford, Ibuprofen: A Critical Bibliographic Review. (Taylor & Francis,
2003).
[137] G. Geisslinger, K. Dietzel, H. Bezler, B. Nuernberg and K. Brune, Int. J. Clin.
Pharmacol. Ther., 27, 324-328 (1989).
[138] V. Legrand, M. Descamps and C. Alba-Simionesco, Thermochim. Acta, 307, 77-
83 (1997).
[139] F. Paladi and M. Oguni, Phys. Rev. B, 65, 144202 (2002).
[140] N. Okamoto and M. Oguni, Solid State Commun., 99, 53-56 (1996).
[141] S. F. Chow, M. Chen, L. M. Shi, A. H. L. Chow and C. C. Sun, Pharm. Res., 29,
1854-1865 (2012).
[142] L. Seton, M. Roberts and F. Ur-Rehman, Chem. Eng. J., 164, 449-452 (2010).
[143] F. Xu, L.-X. Sun, Z.-C. Tan, J.-G. Liang and R.-L. Li, Thermochim. Acta, 412,
33-37 (2004).
[144] T. Sato and C. Sano, Patent No. EP0703214A2 (1995).
[145] CSD version 5.33, November 2011
[146] B. Khawas, Acta Crystallogr. Sect. B, 26, 1919-1922 (1970).
[147] B. Khawas, Ind. J. Phys, 59A, 219-226 (1985).
[148] M. D. King, T. N. Blanton and T. M. Korter, Phys. Chem. Chem. Phys., 14,
1113-1116 (2012).
139
[149] G. A. Stephenson, E. G. Groleau, R. L. Kleemann, W. Xu and D. R. Rigsbee, J.
Pharm. Sci., 87, 536-542 (1998).
[150] P. R. Perrier and S. R. Byrn, J. Org. Chem., 47, 4671-4676 (1982).
[151] S. Petit and G. Coquerel, Chem. Mater., 8, 2247-2258 (1996).
[152] H. Nagase, N. Ogawa, T. Endo, M. Shiro, H. Ueda and M. Sakurai, J. Phys.
Chem. B, 112, 9105-9111 (2008).
[153] M. O. Okoth, R. M. Vrcelj, D. B. Sheen and J. N. Sherwood, CrystEngComm,
14, 1602-1612 (2012).
[154] S. Padmanabhan, S. Suresh and M. Vijayan, Acta Crystallogr. Sect. C, 51, 2098-
2100 (1995).
[155] A. L. Spek, J. Appl. Crystallogr., 36, 7-13 (2003).
[156] A. L. Spek, PLATON. A Multipurpose Crystallographic Tool (Utrecht
University, The Netherlands, 2008).
[157] A. Boultif and D. Louër, J. Appl. Crystallogr., 24, 987-993 (1991).
[158] P. Suortti, J. Appl. Crystallogr., 5, 325-331 (1972).
[159] M. S. Lehmann, J. J. Verbist, W. C. Hamilton and T. F. Koetzle, J. Chem. Soc.,
Perkin Trans. 2, 133-137 (1973).
[160] G. Kartha and A. Devries, Nature, 192, 862-863 (1961).
[161] K. Umadevi, K. Anitha, B. Sridhar, N. Srinivasan and R. K. Rajaram, Acta
Crystallogr. Sect. E, 59, O1073-O1075 (2003).
[162] J. Janczak and P. Luger, Acta Crystallogr. Sect. C, 53, 1954-1956 (1997).
[163] M. N. Frey, M. S. Lehmann, T. F. Koetzle and W. C. Hamilton, Acta
Crystallogr. Sect. B, 29, 876-884 (1973).
[164] S. Chakraborty, A. K. Bera, S. Ghosh, S. Bhattacharya, A. K. Pal, B. P.
Mukhopadhyay and A. Banerjee, Indian J. Phys., A, 75, 499-501 (2001).
[165] S. Suresh, S. Padmanabhan and M. Vijayan, J. Biomol. Struct. Dyn., 11, 1425-
1435 (1994).
[166] R. Kingsford-Adaboh, M. Grosche, B. Dittrich and P. Luger, Acta Crystallogr.
Sect. C, 56, 1274-1276 (2000).
[167] Z. Ciunik and T. Głowiak, Acta Crystallogr. Sect. C, 39, 1271-1273 (1983).
[168] R. O. Fox, Jr. and R. D. Rosenstein, ACS Abstract Papers (Summer), 50 (1976).
140
Appendix A: Atomic Parameters for Crystal
Structures Determined During This Work
x y z Uiso
N1 0.56655(26) 0.1471(7) 0.4808(6) 0.0247(10)
C2 0.61825(24) 0.5932(5) 0.1746(5) 0.0247(10)
C3 0.57946(15) 0.4539(7) 0.2708(6) 0.0247(10)
C4 0.60490(22) 0.2876(6) 0.3832(5) 0.0247(10)
C5 0.66909(24) 0.2604(6) 0.3986(5) 0.0247(10)
C6 0.70786(16) 0.3998(8) 0.3026(7) 0.0247(10)
C7 0.68245(23) 0.5662(7) 0.1905(6) 0.0247(10)
C8 0.59109(22) 0.7715(5) 0.0545(6) 0.0247(10)
O9 0.53318(25) 0.8048(8) 0.0773(11) 0.0247(10)
O10 0.62729(23) 0.8885(8) –0.0712(10) 0.0247(10)
H11 0.5910(7) 0.0555(30) 0.587(12) 0.0346(12)
H12 0.5396(16) 0.1908(23) 0.640(10) 0.0346(12)
H13 0.5434(18) 0.101(5) 0.276(4) 0.0346(12)
H31 0.53433(19) 0.4727(10) 0.2599(8) 0.0346(12)
H51 0.6869(4) 0.1436(8) 0.4776(7) 0.0346(12)
H61 0.75299(20) 0.3810(12) 0.3136(9) 0.0346(12)
H71 0.70974(34) 0.6641(9) 0.1227(7) 0.0346(12)
Table A.1: Atomic coordinates and isotropic displacement parameters (Uiso) for
Form III of m-aminobenzoic acid.
141
x y z Uiso
N1 0.16788(32) 0.25994(33) 0.12903(21) 0.0351(9)
C2 0.35922(31) 0.11399(29) –0.13114(10) 0.0351(9)
C3 0.29725(26) 0.12947(25) –0.03459(9) 0.0351(9)
C4 0.23251(26) 0.24475(26) 0.02797(17) 0.0351(9)
C5 0.22738(28) 0.34552(32) –0.00253(26) 0.0351(9)
C6 0.28940(22) 0.3295(4) –0.09898(29) 0.0351(9)
C7 0.35494(20) 0.2141(4) –0.16277(23) 0.0351(9)
C8 0.43069(26) –0.01103(33) –0.201085(34 0.0351(9)
O9 0.36869(24) –0.10335(31) –0.17453(10) 0.0351(9)
O10 0.5471(4) –0.0164(4) –0.28128(13) 0.0351(9)
H11 0.3794(27) 0.273(7) 0.16557(33) 0.0420(10)
H12 0.035(13) 0.1896(12) 0.1300(21) 0.0420(10)
H13 0.048(17) 0.326(4) 0.1545(12) 0.0420(10)
H31 0.29950(26) 0.06166(26) –0.01208(8) 0.0420(10)
H51 0.1824(4) 0.42396(33) 0.04149(31) 0.0420(10)
H61 0.2870(5) 0.3974(5) –0.1212(5) 0.0420(10)
H71 0.39704(21) 0.2037(4) –0.22841(25) 0.0420(10)
N11 –0.1392(5) 0.7794(5) 0.64344(26) 0.0351(9)
C12 –0.11783(22) 0.60658(35) 0.37564(20) 0.0351(9)
C13 –0.17950(25) 0.63224(33) 0.47311(19) 0.0351(9)
C14 –0.0754(4) 0.75135(34) 0.54053(23) 0.0351(9)
C15 0.0899(4) 0.84638(33) 0.51314(31) 0.0351(9)
C16 0.15090(34) 0.8203(4) 0.41571(33) 0.0351(9)
C17 0.04729(24) 0.7008(4) 0.34730(26) 0.0351(9)
C18 –0.22873(26) 0.4774(4) 0.30060(19) 0.0351(9)
O19 –0.36092(33) 0.39356(35) 0.32978(24) 0.0351(9)
O20 –0.18280(35) 0.4600(5) 0.21176(22) 0.0351(9)
H111 –0.235(17) 0.8518(35) 0.6660(13) 0.0420(10)
H112 0.070(5) 0.788(6) 0.67947(29) 0.0420(10)
H113 –0.293(11) 0.7161(14) 0.6488(22) 0.0420(10)
H131 –0.29207(25) 0.5686(4) 0.49326(20) 0.0420(10)
H151 0.1598(5) 0.92778(34) 0.5604(4) 0.0420(10)
H161 0.2635(4) 0.8841(4) 0.3958(4) 0.0420(10)
H171 0.08956(24) 0.6834(5) 0.28072(27) 0.0420(10)
Table A.2: Atomic coordinates and isotropic displacement parameters (Uiso) for
Form IV of m-aminobenzoic acid.
142
x y z Occ Uiso
N1a 0.81567(35) 0.8863(10) 0.2269(6) 0.602(5) 0.0468(11)
N1b 0.5794(6) 0.9234(11) 0.5392(5) 0.398(5) 0.0468(11)
C2 0.60763(24) 0.50411(34) 0.21252(30) 1.0 0.0468(11)
C3 0.69078(25) 0.5969(5) 0.17642(27) 1.0 0.0468(11)
C4 0.73563(22) 0.7959(6) 0.2608(4) 1.0 0.0468(11)
C5 0.69738(31) 0.9019(4) 0.3812(4) 1.0 0.0468(11)
C6 0.61426(31) 0.8092(5) 0.41732(26) 1.0 0.0468(11)
C7 0.56930(21) 0.6102(6) 0.33296(35) 1.0 0.0468(11)
C8 0.56020(21) 0.2907(4) 0.12113(27) 1.0 0.0468(11)
O9a 0.5977(4) 0.1888(9) 0.01516(32) 0.602(5) 0.0468(11)
O9b 0.48418(22) 0.2109(12) 0.1560(7) 0.398(5) 0.0468(11)
O10a 0.48156(33) 0.2120(10) 0.1534(6) 0.602(5) 0.0468(11)
O10b 0.5992(5) 0.1883(11) 0.0130(5) 0.398(5) 0.0468(11)
H11a 0.8419(18) 0.783(8) 0.1515(33) 0.602(5) 0.0562(13)
H11b 0.572(7) 0.797(6) 0.622(6) 0.398(5) 0.0562(13)
H12a 0.8519(13) 0.950(10) 0.3043(19) 0.602(5) 0.0562(13)
H12b 0.607(5) 1.079(11) 0.567(8) 0.1989(25) 0.0562(13)
H12c 0.523(4) 0.997(28) 0.517(4) 0.1989(25) 0.0562(13)
H31 0.7177(4) 0.5224(8) 0.09185(35) 1.0 0.0562(13)
H41b 0.79407(25) 0.8611(9) 0.2355(6) 0.398(5) 0.0562(13)
H51 0.7289(4) 1.0415(5) 0.4403(5) 1.0 0.0562(13)
H61a 0.5874(4) 0.8838(8) 0.50191(32) 0.602(5) 0.0562(13)
H71 0.51085(25) 0.5449(9) 0.3583(5) 1.0 0.0562(13)
H101a 0.4572(5) 0.0665(10) 0.0845(9) 0.602(5) 0.0562(13)
H101b 0.5612(8) 0.0419(9) –0.0368(6) 0.398(5) 0.0562(13)
Table A.3: Atomic coordinates, occupancy (Occ) and isotropic displacement
parameters (Uiso) for Form V of m-aminobenzoic acid.
143
x y z Uiso
O1 –0.6076(9) 0.1623(18) –0.38044(16) 0.0401(15)
O2 –0.4298(9) 0.1327(22) –0.38981(18) 0.0401(15)
C3 –0.5137(6) 0.2613(12) –0.38219(9) 0.0401(15)
C4 –0.5013(6) 0.5399(13) –0.37506(13) 0.0401(15)
H5 –0.4510(9) 0.5949(18) –0.4193(4) 0.0481(18)
N6 –0.6171(9) 0.6537(18) –0.4086(7) 0.0401(15)
H7 –0.6725(18) 0.566(24) –0.378(11) 0.0481(18)
H8 –0.616(7) 0.828(9) –0.389(11) 0.0481(18)
H9 –0.638(7) 0.641(31) –0.4805(16) 0.0481(18)
C10 –0.4508(7) 0.6117(18) –0.26885(26) 0.0401(15)
H11 –0.3808(5) 0.5199(33) –0.2488(4) 0.0481(18)
H12 –0.4339(14) 0.7886(21) –0.2668(7) 0.0481(18)
C13 –0.5286(5) 0.5570(11) –0.19674(24) 0.0401(15)
C14 –0.6134(6) 0.7194(11) –0.1826(4) 0.0401(15)
C15 –0.6842(5) 0.6680(19) –0.1165(5) 0.0401(15)
C16 –0.6726(7) 0.4521(21) –0.0621(4) 0.0401(15)
C17 –0.5899(9) 0.2871(14) –0.0740(4) 0.0401(15)
C18 –0.5172(6) 0.3379(14) –0.1413(4) 0.0401(15)
H19 –0.6252(10) 0.8722(10) –0.2211(6) 0.0481(18)
H20 –0.7417(5) 0.7876(27) –0.1061(9) 0.0481(18)
H21 –0.7185(10) 0.4247(30) –0.0111(5) 0.0481(18)
H22 –0.5765(13) 0.1417(15) –0.0313(5) 0.0481(18)
H23 –0.4565(7) 0.2229(21) –0.1479(7) 0.0481(18)
O24 –0.8367(10) 0.5974(19) –0.38417(26) 0.0401(15)
O25 –1.0172(9) 0.6181(19) –0.38758(21) 0.0401(15)
C26 –0.9287(7) 0.4942(10) –0.37993(13) 0.0401(15)
C27 –0.9336(5) 0.2167(10) –0.36544(15) 0.0401(15)
H28 –1.0097(6) 0.1576(15) –0.3905(5) 0.0481(18)
N29 –0.8524(9) 0.0930(19) –0.4212(6) 0.0401(15)
H30 –0.7788(11) 0.172(8) –0.404(4) 0.0481(18)
H31 –0.846(4) –0.082(4) –0.403(5) 0.0481(18)
H32 –0.8797(24) 0.109(11) –0.4922(6) 0.0481(18)
C33 –0.9030(7) 0.1591(12) –0.25516(28) 0.0401(15)
H34 –0.8849(11) –0.0166(16) –0.2463(8) 0.0481(18)
H35 –0.8377(6) 0.2577(25) –0.22616(30) 0.0481(18)
Table A.4: Atomic coordinates and isotropic displacement parameters (Uiso) for
L-phenylalanine hemihydrate (continued on next page).
144
C36 –1.0007(6) 0.2204(10) –0.20436(25) 0.0401(15)
C37 –1.0957(7) 0.0719(11) –0.21524(33) 0.0401(15)
C38 –1.1842(6) 0.1307(16) –0.1683(5) 0.0401(15)
C39 –1.1792(7) 0.3392(18) –0.1095(5) 0.0401(15)
C40 –1.0860(9) 0.4872(12) –0.09814(33) 0.0401(15)
C41 –0.9967(7) 0.4287(13) –0.1453(4) 0.0401(15)
H42 –1.1012(11) –0.0741(10) –0.2578(4) 0.0481(18)
H43 –1.2497(7) 0.0219(23) –0.1747(7) 0.0481(18)
H44 –1.2390(10) 0.3725(25) –0.0716(6) 0.0481(18)
H45 –1.0792(13) 0.6264(13) –0.0520(4) 0.0481(18)
H46 –0.9292(8) 0.5319(20) –0.1352(6) 0.0481(18)
O47 –1.2420(15) 0.4368(32) –0.4037(12) 0.071(8)
Table A.4 (cont.)
145
x y z Occ Uiso
O1 0.3528(7) 0.1213(17) 0.63877(23) 1.0 0.0671(15)
O2 0.5169(8) 0.0620(20) 0.65024(24) 1.0 0.0671(15)
C3 0.4430(5) 0.2040(12) 0.64641(14) 1.0 0.0671(15)
C4 0.4635(4) 0.4799(13) 0.65113(16) 1.0 0.0671(15)
H5 0.5117(7) 0.5212(17) 0.6074(4) 1.0 0.0805(19)
N6 0.3612(8) 0.6105(19) 0.6168(6) 1.0 0.0671(15)
H7 0.3075(18) 0.540(15) 0.649(6) 1.0 0.0805(19)
H8 0.370(4) 0.786(4) 0.633(7) 1.0 0.0805(19)
H9 0.340(5) 0.591(18) 0.5458(13) 1.0 0.0805(19)
C10 0.5132(6) 0.5558(16) 0.75576(28) 1.0 0.0671(15)
H11 0.5782(5) 0.4627(30) 0.7765(4) 1.0 0.0805(19)
H12 0.5299(12) 0.7318(19) 0.7564(7) 1.0 0.0805(19)
C13 0.4417(5) 0.5085(12) 0.82739(25) 1.0 0.0671(15)
C14 0.4535(5) 0.2966(11) 0.88460(34) 1.0 0.0671(15)
C15 0.3889(7) 0.2508(13) 0.95043(29) 1.0 0.0671(15)
C16 0.3102(6) 0.4156(20) 0.9612(4) 1.0 0.0671(15)
C17 0.2964(6) 0.6274(19) 0.9056(6) 1.0 0.0671(15)
C18 0.3619(7) 0.6756(13) 0.8384(4) 1.0 0.0671(15)
H19 0.5083(7) 0.1771(15) 0.8785(6) 1.0 0.0805(19)
H20 0.3983(11) 0.1006(15) 0.99018(33) 1.0 0.0805(19)
H21 0.2649(8) 0.3812(30) 1.0082(6) 1.0 0.0805(19)
H22 0.2410(7) 0.7443(25) 0.9131(9) 1.0 0.0805(19)
H23 0.3521(11) 0.8261(14) 0.7989(5) 1.0 0.0805(19)
O24 –0.0168(8) 0.5976(18) 0.61102(26) 1.0 0.0671(15)
O25 0.1504(8) 0.5639(18) 0.62153(27) 1.0 0.0671(15)
C26 0.0633(6) 0.4686(10) 0.62439(15) 1.0 0.0671(15)
C27 0.0550(5) 0.1946(10) 0.64418(18) 1.0 0.0671(15)
H28 –0.0166(5) 0.1390(15) 0.6184(5) 1.0 0.0805(19)
N29 0.1283(7) 0.0675(18) 0.5910(6) 1.0 0.0671(15)
H30 0.0997(20) 0.072(9) 0.5206(6) 1.0 0.0805(19)
H31 0.1963(11) 0.151(6) 0.604(4) 1.0 0.0805(19)
H32 0.1371(30) –0.1040(34) 0.6129(35) 1.0 0.0805(19)
C33 0.0827(6) 0.1396(11) 0.75400(31) 1.0 0.0671(15)
H34 0.0941(11) –0.0377(13) 0.7641(8) 1.0 0.0805(19)
H35 0.1466(6) 0.2289(24) 0.78310(32) 1.0 0.0805(19)
Table A.5: Atomic coordinates, occupancy (Occ) and isotropic displacement
parameters (Uiso) for L-phenylalanine monohydrate (continued on next page).
146
C36 –0.0065(6) 0.2215(9) 0.80171(27) 1.0 0.0671(15)
C37 –0.1018(7) 0.0974(10) 0.78349(31) 1.0 0.0671(15)
C38 –0.1834(5) 0.1717(15) 0.8269(5) 1.0 0.0671(15)
C39 –0.1715(7) 0.3705(16) 0.8890(5) 1.0 0.0671(15)
C40 –0.0780(9) 0.4968(11) 0.9083(4) 1.0 0.0671(15)
C41 0.0048(6) 0.4252(13) 0.8656(4) 1.0 0.0671(15)
H42 –0.1115(10) –0.0453(10) 0.73962(32) 1.0 0.0805(19)
H43 –0.2504(6) 0.0834(22) 0.8131(7) 1.0 0.0805(19)
H44 –0.2301(10) 0.4248(23) 0.9187(7) 1.0 0.0805(19)
H45 –0.0701(14) 0.6405(12) 0.9517(4) 1.0 0.0805(19)
H46 0.0710(7) 0.5168(21) 0.8794(6) 1.0 0.0805(19)
O47 0.7690(8) 0.5281(27) 0.6098(11) 1.0 0.074(7)
H48 0.739(16) 0.688(15) 0.587(6) 0.5 0.089(8)
H49 0.8450(23) 0.55(4) 0.626(4) 1.0 0.089(8)
O50 0.3103(9) 0.5127(30) 0.4167(17) 1.0 0.064(6)
H51 0.3811(27) 0.524(26) 0.403(6) 1.0 0.077(7)
H52 0.280(10) 0.679(13) 0.407(29) 0.5 0.077(7)
H53 0.759(15) 0.355(17) 0.590(20) 0.5 0.089(8)
H54 0.291(10) 0.338(8) 0.411(21) 0.5 0.077(7)
Table A.5 (cont.)
147
x y z Uiso
O1 –0.6130(6) 0.1690(15) –0.38532(13) 0.087(6)
O2 –0.4369(6) 0.1233(18) –0.39033(14) 0.087(6)
C3 –0.5182(5) 0.2601(12) –0.38481(7) 0.087(6)
C4 –0.5006(5) 0.5381(12) –0.37752(10) 0.087(6)
H5 –0.4490(7) 0.5880(16) –0.42214(28) 0.105(7)
N6 –0.6123(7) 0.6633(17) –0.4092(5) 0.087(6)
H7 –0.6713(14) 0.569(20) –0.384(11) 0.105(7)
H8 –0.609(7) 0.831(11) –0.382(10) 0.105(7)
H9 –0.629(7) 0.671(28) –0.4817(6) 0.105(7)
C10 –0.4503(5) 0.6111(14) –0.27103(20) 0.087(6)
H11 –0.3795(4) 0.5217(24) –0.25122(34) 0.105(7)
H12 –0.4349(10) 0.7890(16) –0.2689(5) 0.105(7)
C13 –0.5286(4) 0.5519(10) –0.19855(19) 0.087(6)
C14 –0.6127(5) 0.7162(10) –0.18332(33) 0.087(6)
C15 –0.6840(4) 0.6619(15) –0.1171(4) 0.087(6)
C16 –0.6732(5) 0.4416(17) –0.06431(34) 0.087(6)
C17 –0.5909(6) 0.2749(12) –0.07755(32) 0.087(6)
C18 –0.5182(4) 0.3266(12) –0.14399(35) 0.087(6)
H19 –0.6240(8) 0.8694(10) –0.2219(5) 0.105(7)
H20 –0.7399(4) 0.7835(21) –0.1041(7) 0.105(7)
H21 –0.7152(7) 0.4165(23) –0.0097(4) 0.105(7)
H22 –0.5748(9) 0.1338(12) –0.0325(4) 0.105(7)
H23 –0.4565(5) 0.2129(16) –0.1501(6) 0.105(7)
O24 –0.8261(7) 0.5891(14) –0.38431(19) 0.087(6)
O25 –1.0063(6) 0.6426(15) –0.39248(16) 0.087(6)
C26 –0.9230(5) 0.5022(8) –0.38181(10) 0.087(6)
C27 –0.93804(32) 0.2270(8) –0.36585(11) 0.087(6)
H28 –1.0159(4) 0.1798(11) –0.3916(4) 0.105(7)
N29 –0.8600(5) 0.0880(17) –0.4205(5) 0.087(6)
H30 –0.7837(8) 0.155(7) –0.4027(35) 0.105(7)
H31 –0.8600(34) –0.0873(28) –0.403(4) 0.105(7)
H32 –0.8851(26) 0.106(9) –0.4921(4) 0.105(7)
C33 –0.9099(5) 0.1608(10) –0.25567(21) 0.087(6)
H34 –0.8969(9) –0.0176(12) –0.2488(6) 0.105(7)
H35 –0.8417(5) 0.2491(20) –0.22569(24) 0.105(7)
Table A.6: Atomic coordinates and isotropic displacement parameters (Uiso) for
polymorph II of L-phenylalanine (continued on next page).
148
C36 –1.0057(4) 0.2324(9) –0.20306(20) 0.087(6)
C37 –1.1007(5) 0.0836(10) –0.21167(30) 0.087(6)
C38 –1.1884(4) 0.1483(15) –0.1637(4) 0.087(6)
C39 –1.1816(6) 0.3621(16) –0.1070(4) 0.087(6)
C40 –1.0871(7) 0.5103(12) –0.09837(29) 0.087(6)
C41 –0.9992(5) 0.4467(11) –0.14605(30) 0.087(6)
H42 –1.1067(8) –0.0635(10) –0.25406(34) 0.105(7)
H43 –1.2535(5) 0.0384(21) –0.1665(6) 0.105(7)
H44 –1.2376(8) 0.3947(22) –0.0643(5) 0.105(7)
H45 –1.0768(10) 0.6455(12) –0.05002(32) 0.105(7)
Table A.6 (cont.)
149
Appendix B: Full Differential Scanning Calorimetry
Thermograms for Experiments Performed on
rac-Ibuprofen
Figure B.1: DSC data for Experiment A. Mass of sample = 4.9 mg.
150
Figure B.2: DSC data for Experiment B. Mass of sample = 3.5 mg.
Figure B.3: DSC data for Experiment C. Mass of sample = 3.4 mg.
151
Figure B.4: DSC data for Experiment D. Mass of sample = 3.6 mg.
Figure B.5: DSC data for Experiment E. Mass of sample = 3.7 mg.
152
Figure B.6: DSC data for Experiment F. Mass of sample = 3.4 mg.
Figure B.7: DSC data for Experiment G. Mass of sample = 3.9 mg.
153
Figure B.8: DSC data for Experiment H. Mass of sample = 3.2 mg.
Figure B.9: DSC data for Experiment I. Mass of sample = 3.4 mg.
154
Figure B.10: DSC data for Experiment J. Mass of sample = 3.5 mg.
Figure B.11: DSC data for Experiment K. Mass of sample = 3.7 mg.
155
Figure B.12: DSC data for Experiment L. Mass of sample = 3.7 mg.
Figure B.13: DSC data for Experiment M. Mass of sample = 3.7 mg.
156
Figure B.14: DSC data for Experiment N. Mass of sample = 3.5 mg.
157
Appendix C: Work Published during this PhD
E. Courvoisier, P. A. Williams, G. K. Lim, C. E. Hughes and K. D. M. Harris; The
Crystal Structure of L-Arginine; Chem. Commun., 48, 2761-2763 (2012).
P. A. Williams, C. E. Hughes, G. K. Lim, B. M. Kariuki and K. D. M. Harris; Discovery
of a New System Exhibiting Abundant Polymorphism: m-Aminobenzoic Acid; Cryst.
Growth Des., 12, 3104-3113 (2012).
C. E. Hughes, P. A. Williams, T. R. Peskett and K. D. M. Harris; Exploiting In Situ
Solid-State NMR for the Discovery of New Polymorphs during Crystallization
Processes; J. Phys. Chem. Lett., 3, 3176-3181 (2012).
P. A. Williams, C. E. Hughes and K. D. M. Harris; New Insights into the Preparation of
the Low-Melting Polymorph of Racemic Ibuprofen; Cryst. Growth Des., 12, 5839-5845
(2012).
D. V. Dudenko, P. A. Williams, C. E. Hughes, O. N. Antzutkin, S. P. Velaga, S. P.
Brown and K. D. M. Harris; Exploiting the Synergy of Powder X-ray Diffraction and
Solid-State NMR Spectroscopy in Structure Determination of Organic Molecular
Solids; J. Phys. Chem. C, 117, 12258-12265 (2013).
P. A. Williams, C. E. Hughes, A. B. M. Buanz, S. Gaisford and K. D. M. Harris;
Expanding the Solid-State Landscape of L-Phenylalanine: Discovery of Polymorphism
and New Hydrate Phases, with Rationalization of Hydration/Dehydration Processes; J.
Phys. Chem. C, 117, 12136-12145 (2013).
K. D. M. Harris and P. A. Williams; Structure from Diffraction Methods: Inorganic
Materials Series (Eds.: D. W. Bruce, D. O’Hare, R. I. Walton); Chapter 1: Powder
Diffraction, Pg. 1-81 (2014).
S. Iwama, K. Kuyama, Y. Mori, K. Manoj, R. G. Gonnade, K. Suzuki, C. E. Hughes,
P. A. Williams, K. D. M. Harris, S. Veesler, H. Takahashi, H. Tsue and R. Tamura;
Highly Efficient Chiral Resolution of DL-Arginine by Cocrystal Formation Followed by
Recrystallization under Preferential Enrichment Conditions; Chem. Eur. J., 20,
10343-10350 (2014).
158
C. E. Hughes, P. A. Williams and K. D. M. Harris; “CLASSIC NMR”: An In-Situ NMR
Strategy for Mapping the Time-Evolution of Crystallization Processes by Combined
Liquid-State and Solid-State Measurements; Angew. Chem. Int. Ed., 53, 8939-8943
(2014).